conceptualizing agency: folkpsychological and ... agency: folkpsychological and folkcommunicative...

Cognition 162 (2017) 103–123

Contents lists available at ScienceDirect

Cognition

journal homepage: www.elsevier .com/locate /COGNIT

Original Articles

Conceptualizing agency: Folkpsychological and folkcommunicativeperspectives on plants

http://dx.doi.org/10.1016/j.cognition.2017.01.0230010-0277/� 2017 Elsevier B.V. All rights reserved.

⇑ Corresponding author.E-mail address: [email protected] (b. l. ojalehto).

bethany l. ojalehto a,⇑, Douglas L. Medin a, Salino G. García b

a Psychology Department, Northwestern University, 2029 Sheridan Road – 102 Swift Hall, Evanston, IL 60208-2710, United StatesbNgöbe Culture and Language Education Program, Bocas del Toro, Panama

a r t i c l e i n f o

Article history:Received 2 September 2015Revised 28 January 2017Accepted 31 January 2017

Keywords:Agency conceptsCultureFolkpsychologyFolkcommunicationIndigenous

a b s t r a c t

The present research addresses cultural variation in concepts of agency. Across two experiments, weinvestigate how Indigenous Ngöbe of Panama and US college students interpret and make inferencesabout nonhuman agency, focusing on plants as a critical test case. In Experiment 1, participants predictedgoal-directed actions for plants and other nonhuman kinds and judged their capacities for intentionalagency. Goal-directed action is pervasive among living kinds and as such we expected cultural agreementon these predictions. However, we expected that interpretation of the capacities involved would differbased on cultural folktheories. As expected, Ngöbe and US participants both inferred that plants wouldengage in goal-directed action but Ngöbe were more likely to attribute intentional agency capacities toplants. Experiment 2 extends these findings by investigating action predictions and capacity attributionslinked to complex forms of plant social agency recently discovered in botanical sciences (communication,kin altruism). We hypothesized that the Ngöbe view of plants as active agents would productively guideinferences for plant social interaction. Indeed, Ngöbe were more likely than US participants to infer thatplants can engage in social behaviors and they also attributed more social agency capacities to plants. Weconsolidate these findings by using bottom-up consensus modeling to show that these cultural differ-ences reflect two distinct conceptual models of agency rather than variations on a single (universal)model. We consider these findings in light of current theories of domain-specificity and animism, andoffer an alternative account based on a folktheory of communication that infers agency on the basis ofrelational interactions rather than having a mind.

� 2017 Elsevier B.V. All rights reserved.

1. Introduction

A long-standing question in psychology concerns how peoplerecognize and reason about agents—the kinds of things that arecapable of acting in, about, and towards the world. It is widely pro-posed that this rests on a dualistic distinction between psycholog-ical agents versus insentient actors (Johnson, 2003; Wellman &Johnson, 2008). Specifically, the actions of psychological agents likehumans or other animals are explained in terms of mental stateslike beliefs and desires—a conceptual framework known asfolkpsychology (e.g., Erickson, Keil, & Lockhart, 2010). By contrast,the actions of other kinds of actors like plants or clouds areexplained by appeal to non-intentional biological (folkbiology) orphysical forces (folkphysics). For instance, people purportedly acti-vate different folktheories to interpret what causes movementtoward the sun by a cloud versus a plant (folkphysics versus

folkbiology), and both would be interpreted differently from whatcauses a person to move toward the sun (folkpsychology) (Gutheil,Vera, & Keil, 1998; Opfer & Gelman, 2001; Opfer & Siegler, 2004).This partitioning of domains presents a worldview in which enti-ties in the natural world can be adequately described in terms ofphysical and biological causes without recourse to intentionalityor sociality—properties that are considered exclusive to humansand perhaps some other animals. The corresponding theory offolkpsychology represents an anthropocentric model that mayextend agency to some ‘‘higher” animals such as mammals, butexcludes entities such as plants, minerals, or water bodies on thegrounds that they do not share similar forms of agency to humans(e.g., Carey, 1985).

This is the account proposed by domain-specificity theory,which sees these intuitive folktheories as fundamental, untutoredways of organizing knowledge about the world that reflect univer-sal cognitive structure (Hirschfeld & Gelman, 1994; Sperber &Hirschfeld, 2004; Wellman & Gelman, 1992). However, thisparticular delineation of intuitive domains has been formulated

http://crossmark.crossref.org/dialog/?doi=10.1016/j.cognition.2017.01.023&domain=pdf

http://dx.doi.org/10.1016/j.cognition.2017.01.023

mailto:[email protected]


http://www.sciencedirect.com/science/journal/00100277

http://www.elsevier.com/locate/COGNIT

104 b. l. ojalehto et al. / Cognition 162 (2017) 103–123

byWestern psychologists and much of the evidence in support of ithas come from Western study populations (Carey, 2009; Gutheilet al., 1998; Keil, 1995; Opfer, 2002; Opfer & Gelman, 2001), andsome industrialized Eastern cultures (Hatano & Inagaki, 1994;Inagaki & Hatano, 2004). Yet more recent research suggests thatthis organization of domains may reflect Western cultural episte-mologies. For instance, the Western anthropocentric stance leadsto patterns of folkbiological reasoning that differ markedly fromthose of Indigenous communities that take an ecological stance(Herrmann, Waxman, & Medin, 2010; Ross, Medin, Coley, &Atran, 2003; Unsworth et al., 2012; Waxman & Medin, 2007).Based on such findings, it has been proposed that alternative par-titionings of domains may provide a better match for conceptualpatterns in different cultures—for example, folkecology may bemore apt than folkbiology in some Indigenous communities(ojalehto & Medin, 2015). Similarly, it is a distinct but untestedpossibility that the anthropocentric stance in folkpsychology maylikewise reflect a Western orientation that is not shared in othercultures. In the current research, we propose that folkcommunica-tion may be more apt than folkpsychology in the Indigenous Ngöbecommunity of Panama. The present experiments investigatewhether Indigenous Ngöbe adults and US college students hold dif-ferent conceptual frameworks for agency that facilitate distinctsets of inferences about and interpretations of nonhuman agency,focusing on plants as a critical test case.1

2. Background

2.1. Indigenous perspectives on agency

Anthropological observations suggest that many Indigenouscommunities organize knowledge about agents in ways that differfrom Western folkpsychology (Lillard, 1998; Luhrmann, 2011).First, understanding of others is framed more in terms of relationalinteraction or communication than individual subjectivity (Ingold,2010). Here the focus is on explaining actions in terms of overtbehavior, relationships and social roles, and speech acts or othersigns rather than explicit inferences about the private mentalstates of others (Danziger, 2006, 2010; Danziger & Rumsey, 2013;Duranti, 1988, 2008; Robbins & Rumsey, 2008). Second, communi-cation with nonhumans is a significant concern in communitieswhere everyday life involves intimate interaction with ecologies(Harvey, 2005; Ingold, 2006, 2011; Kohn, 2013). Personally navi-gating relationships with nonhuman kinds such as soil, thunder,plants, or animals requires close attention to nonhuman behaviorsand responses (Anderson, 1996, 2012; Hallowell, 1960; Rappaport,1979; Reichel-Dolmatoff, 1976). Native and non-native scholarsalike have argued that these exchanges are founded on a concep-tion of nonhumans as beings with a perspective or locus of agency,where agency is understood to be expressed through the capacityto participate in relationships (e.g., predator-prey relations, healingrelations) (Bird-David, 1999; Cajete, 2000; Descola, 1996;Kimmerer, 2013; Lima, 1999; Pierotti, 2011; Santos-Granero,2009; Viveiros de Castro, 1998) (see also Ramos, 2012).2

As a consequence of these views on agency, nonhumans areoften included in social relationships within Indigenous communi-

1 On Western folkpsychology, capacities like ‘‘want” and ‘‘think” are interpreted asmental states associated with minds, but it may be that Ngöbe interpret thesecapacities differently. Thus, we refer broadly to ‘‘agency” rather than ‘‘minds” or‘‘mental states” in order to avoid specific folkpsychological connotations.

2 These observations have been theorized under distinct frameworks, includinganimism, perspectivism, and relational epistemology. There are significant differencesamong these approaches and among Indigenous communities (e.g., Ramos, 2012).However, while anthropological theories of human-nonhuman interaction aredebated, the significance of such interactions in everyday life among Indigenouscommunities is not (Ramos, 2012, p. 483).

ties. For instance, dogs are administered medicinal treatments toenhance their perceptual acuity while hunting, and their behaviors,from barking (while awake or dreaming) to body language and eyecontact, are analyzed as meaningful forms of communication (e.g.,Kohn, 2007). Plants and animals are also included in the moral cir-cle, as exemplified by such customs as prohibiting rude actions(e.g., spitting) toward plants (Santos-Granero, 2012) and chastisingchildren for wanton harm to insects (Anderson, 2005).

In sum, many Indigenous communities emphasize attention tononhuman agency and communication—two principles thatanthropologists have argued diverge from the Western focus onhumans and psychology. To be sure, there is great diversity amongIndigenous communities (Descola, 1996). Yet within this diversity,a common principle of many Indigenous epistemologies is that ani-mals, plants, and other natural forces are agents capable of relatingand communicating with others and their environments (Pierotti,2011). In fact, the widespread salience of these principles acrossso many native societies has generated important debates aboutthe nature and extent of cultural variability in this cognitivedomain.

2.2. Understanding cultural variability in agency concepts

Indigenous ideas about nonhuman agency have figured promi-nently in theories of cognitive variability across cultures (e.g.,Guthrie et al., 1980). Several influential accounts have treated theattribution of agency to nonhuman kinds such as plants as a mis-taken inference. Building on the assumption that Western folkpsy-chology is a universal framework that applies to the proper domainof animates (i.e., humans and animals), these accounts argue thatattributing agency to plants represents a ‘‘category error” byextending mentalistic attributions beyond the proper intuitivedomain (Atran & Norenzayan, 2004; Boyer, 1996). Such ‘‘overex-tensions” imply a conceptual slippage that requires explanation,and scholars have argued that they represent counterintuitive con-cepts, religious beliefs, or a strategy adopted in the absence ofknowledge (e.g., mistakenly assuming that plants have minds)(e.g., Barrett, 2000; Boyer, 2003; for similar arguments in thedevelopmental domain, see also Kelemen, 2003). In short, overex-tension accounts assume that the Western folkpsychological con-cept of mind is universal and cultures vary in how far theyextend the concept to nonhumans, resulting in incorrectinferences.

An alternative approach would be to remain neutral about theepistemological status of agency attributions to nonhumans andinstead ask on what conceptual grounds such a framework couldsupport useful inferences. Such inferences would presumably bebased on a folktheory that provides a framework for understandingcomplex behaviors of diverse nonhuman kinds, unlike folkpsychol-ogy which (by definition) is not applicable to plants and other non-animals. One possible alternative could take the form of a folkthe-ory of communication where the relevant explanatory constructsrevolve around relationships and interactions rather than mentalstates. By focusing on mental states, Westerners take an anthro-pocentric stance on the psychological causes of action, presumablyusing knowledge about the kinds of entities that have a brain ornervous system as a fundamental constraint on folkpsychology.Ngöbe do not share this constraint, which frees them to recognizeplant behavior as an instance of agency. A folktheory that usescommunicative principles as the basis for inferring agency wouldprovide a framework for viewing both plants and animals (as wellas other kinds) as agents actively relating to their environments. Byhypothesis, this approach would also encourage a broader stancefor observing and expecting complex plant behaviors, specificallyforms of interacting and relating. From a folkcommunicationperspective, recognition of nonhuman agency need not index a

b. l. ojalehto et al. / Cognition 162 (2017) 103–123 105

misguided attribution of mind, but could instead represent anempirically grounded observation of agency as it is expressedthrough relationships.

The ethnographic record suggests that such a folktheory ofcommunication is evident in various Indigenous communities(e.g., Harvey, 2005; Kohn, 2013), and we have found evidence forthis among Ngöbe individuals (ojalehto, Medin, & Garcia, 2015).When explaining why they attribute agency to plants and othernatural abiotic kinds (e.g., sun, ocean), Ngöbe tend to construeagency as a relational property by referring to observable interac-tions (e.g., growing toward sunlight) and relationships (e.g., pro-ducing seedlings). By contrast, we found that US participantstend to construe agency as a psychological property characterizedby unobservable mental states underlying actions. These findingssuggest that Ngöbe and US individuals use different conceptualframeworks for interpreting the actions of agents, and the Ngöbeframework might be better characterized in terms of folkcommu-nication than folkpsychology.

The above study, like prior ethnographic and psychologicalstudies of animism, relied on verbal reports and predicate attribu-tions (e.g., ‘‘Can [entity] think?”). Assessing whether Ngöbe frame-works lead to productive inferences requires other measures thatcan assess predictive reasoning. To do so we draw on methodsby Opfer and Gelman (2001). They tested the development offolkpsychology and folkbiology by asking US children and adults(1) to predict whether different nonhuman kinds (animals, plants,machines, artifacts) could engage in goal-directed actions, andthen (2) to judge the psychological capacities of each kind. Thisdesign leverages converging measures to highlight the interactionbetween inferential reasoning (e.g., predicting that a plant willengage in communicative behavior) and interpretive frameworks(i.e., saying that a plant can communicate). Both forms of evidenceare critical for assessing the conceptual commitments underlyingnonhuman agency attributions.

3. Current research

The present research investigates whether Ngöbe adults and UScollege students hold distinct folktheories of agency that invokedifferent interpretations of and guide distinct inferences aboutnonhuman agents, specifically plants. Plants are a good test casefor several reasons. First, plants exhibit the hallmark action pat-terns that cue agency perception beginning in infancy (Johnson,2003), including goal-directedness (e.g., phototropism) and contin-gent interaction (e.g., using chemical defenses when attacked byinsects).3 Plants are also seen as living kinds on both Western andIndigenous epistemologies, unlike other natural kinds such as thesun or ocean which Ngöbe but not US participants recognize as alive(ojalehto, Medin, & Garcia, submitted for publication). This providescommon ground for recognizing plant movements as cues to auton-omous agency, even if Westerners use folkbiology rather thanfolkpsychology to interpret plant agency (e.g., Opfer & Gelman,2001).

Second, the nature and development of conceptual knowledgesurrounding plants and plant-human interactions has recentlyemerged as a topic of inquiry in cognitive psychology (Wertz &Wynn, 2014a, 2014b). At the same time, biologists are introducingthe question of plant intelligence to psychology (Abramson &Chicas-Mosier, 2016; Cvrcková, Zársky, & Markoš, 2016;Marder, 2012; Trewavas, 2016). Constructs that are central topsychological theories of domain-specific conceptual organization,such as agency or intelligence, are increasingly complicated by the

3 One obvious difference between plant and animal action patterns is the slowertime scale for sessile plants than motile organisms (in most cases). Apprehendingplant agency may require close observation.

documentation of sophisticated plant behaviors that challengeanthropocentric definitions of those terms.

Finally, plants occupy a contested category in Western scienceand philosophy (e.g., Hall, 2011). Traditionally, Western scholarshave viewed plants as passive objects rather than active agentsand favored explanations of plant activity in terms of automaticreactions and mechanistic processes (Hall, 2011; Trewavas,2002). For instance, Darwin’s claim that plant phototropism is aninductive behavior (not a mechanical process) on the part of plantswas hotly contested among botanists in the 1800s (see vonWiesner (1878) (cf Chamovitz, 2012; Whippo & Hangarter, 2006).Contemporary debates in plant science center on recent findingsthat demonstrate complex social capacities of plants, includingcommunication and kin resource sharing (Bhatt, Khandelwal, &Dudley, 2011; Biedrzycki & Bais, 2010; Brenner et al., 2006;Dudley & File, 2007). These findings were unexpected given theprevailing view of plants as having limited capacities. Yet there isnow mounting evidence to suggest that basic forms of awareness,memory, and sensory perception are present among plants—in-deed, some scientists argue that plant behavior cannot be under-stood without appealing to these capacities (Chamovitz, 2012;Trewavas, 2002, 2003).4 Nonetheless, the evidence remains highlydebated and some plant scientists flat out reject the possibility ofplant sentience (Alpi et al., 2007). It is plausible that these debatesare tied to broader cultural conceptions of plants as passive organ-isms in contrast to sentient animal agents. If so, this would provideindirect evidence to support our hypothesis that interpretativeframeworks for agency interact with expectations for actual behav-ior. On this hypothesis, Ngöbe conceptions of plant capacities maymatch recent botanical science better than do US college studentintuitions. This provides a test bed for observing the role that cul-tural frameworks play in agency concepts, one that we exploit inthe present studies.

The aim of the present research is to explore whether culturalframeworks offer different interpretations for plant capacitiesand how these frameworks guide predictions for actual plantbehavior. Each of our two experiments assessed (a) how peoplemake inferences about behavior (action predictions), and (b) howthey interpret that behavior (capacity attributions). These converg-ing measures allow us to test whether conceptual differences gobeyond interpretive differences to correlate with distinct infer-ences about behavioral patterns. We predicted that both groupswould predict basic forms of goal-directed action among plants(Exp. 1) but that Ngöbe would be more likely than US participantsto predict complex social behaviors among plants (Exp. 2). We alsopredicted that interpretive frameworks would differ such thatNgöbe would be more likely than US participants to attributeintentional agency capacities (i.e., those associated with goal-directed behavior in Exp. 1) and social agency capacities (i.e., thoseassociated with social behavior in Exp. 2) to plants.

Our proposal implies that agency predicates (e.g., want, com-municate) acquire different meanings depending on one’s culturalfolktheory of agency. A folk-communicative grounding for agencyis supported by our prior experiments as well as anthropologicalreports. But we will also need to consider an alternative interpre-tation: one might argue that instead of having a different frame-work for agency, Ngöbe interpret agency predicates in apsychological sense just as Westerners do but differ by extendingthem to plants. Alternatively, it may also be argued that Ngöbehave greater ecological expertise than US college students, whichinforms their knowledge about plant behavior independently of

4 In this regard, it is interesting that some early psychologists argued for a view ofpsychology that would be broad enough to encompass sessile forms of agency(Heidelberger, 2004; Johnson, 1945; Lindsay, 1876).


underlying agency concepts. We consider both possibilities furtherin the general discussion.

3.1. Language

The distinct languages across our cultural samples introducethe possibility of cross-linguistic variation in agency concepts. Eng-lish is the primary language among US college students. In ourPanama research community, most Ngöbe are bilingual in Spanishand Ngöbere, a Chibchan language currently spoken by an esti-mated 133,000 people in Panama and Costa Rica (Lewis, 2009). Asubstantial portion of everyday conversation in the community isconducted in Ngöbere, but Spanish is also used frequently and par-ticularly in conversation with non-Ngöbe individuals. With respectto these various languages, one concern is that differences inagency attribution may reflect different semantic mappings foragency terms across English, Spanish, and/or Ngöbere. As anexploratory analysis of potential language effects, we included lan-guage as a factor within each cultural group (US: Spanish and Eng-lish; Panama: Spanish and Ngöbere).

Previous research points to both universals and variability incross-linguistic constructions of agency. Some researchers haveargued that universal semantic primitives underlie basic agencyterms (e.g., think, feel, want, know, say) across languages, includ-ing Spanish and English (Goddard, 2010). This research has focusedon agency terms as they apply to human subjects, however, so it isunclear if these semantic primitives also extend (or not) to nonhu-man subjects similarly across languages. Furthermore, subtle lin-guistic differences surrounding these putatively universalprimitives could still have psychological consequences (Goddard,2003). Additionally, different grammatical constructions of voli-tional agency in Spanish versus English have been linked to differ-ential memory for accidental agents (Fausey & Boroditsky, 2011;Fausey, Long, Inamori, & Boroditsky, 2010).5 These latter findingssuggest that language could be a factor in agency attributions.

There is no comparable linguistic work on the semantic map-ping of agency terms in Ngöbere (P. Young, personal communica-tion, April 25, 2013), but the basic inventory of mental-stateterms in Ngöbere has been documented (Kopesec, 1975; Reid,1994; Sarsaneda del Cid, 2009). These terms correspond to thoseused in the current protocol with adjustment for regional dialect.Notably, Ngöbere has a distinct classifier system for the categoryof living plants, which is uncommon among world languages andunderscores the attention paid to the vegetal world in Ngöbe cul-ture (Young, 1986).

4. Methods and study populations

Participants in both two-part experiments were drawn from thesame samples of 23 Ngöbe adults (7 female) living in an Indigenouscommunity in Panama and 34 US college students (16 female)attending a private Midwestern university, as described below.

4.1. Panama

4.1.1. Ngöbe populationThe Ngöbe people have resided for millennia in what is now

Panama. The primary research community is located on a forestedisland off the Caribbean coast and numbers around 600 people.Here individuals live in a small-scale community and personallyinteract with ecosystems through a subsistence relationship with

5 To the extent that the effects observed by Fausey et al. might be relevant to thecurrent tasks, they would work against our hypothesis by predicting less agencyattribution to actors (and specifically non-volitional actors) among Spanish speakersthan English speakers.

the land and sea, including agroforestry, fishing, and diving. Dailylife is closely coordinated with the surrounding ecology (Gordon,1982; Johannessen, 1970). In terms of formal institutions, the localpublic school offers primary and partial secondary educationtaught in Spanish based on a westernized curriculum. The commu-nity also hosts two Christian churches and an Indigenous nativisticchurch (Young, 1975).

Ethnographic literature and participant observation indicatesthat the Ngöbe church teaches Indigenous philosophy and prac-tices and also represents a cultural revitalization movement(Young, 1975). Church meetings are conducted in Ngöbere andmany traditional oral histories are maintained in this context(see also Montezuma, 1991; Reid, 1994; Sarsaneda del Cid,2009). Religious practice includes ecological observances, such ascelebrations of lunar events and harvest rituals. Thus, attendanceat the Ngöbe Church can be seen as one index of an individual’sbroader alignment with Ngöbe cultural epistemologies, includingways of knowing, practices, and values.

4.1.2. Ngöbe sampleAges ranged from 18 to 60 years (M = 34.91; SD = 12.89). Partic-

ipants reported 0–13 years of formal schooling (M = 5.74,SD = 3.26). Our sample included 13 self-reported members of thenativistic Ngöbe church, 6 Evangelical church members, and 4non-churchgoers. Ngöbe adults were recruited through personalvisits or word of mouth. Each participant received a small compen-sation and a donation was made to the community fund for theirparticipation. Community permission and individual informedconsent were received.

4.2. United States

4.2.1. US populationUS participants were part of a subject pool associated with an

introductory psychology class at Northwestern University. Stu-dents came from predominantly upper-middle class backgrounds(69%); and students identified as Caucasian (55%), Asian/Asian-American (21%), Black/African-American (5%), Latino/Latina (6%)or other/multiracial (13%).

4.2.2. US sampleAges ranged from 18 to 21 years (M = 18.8, SD = 0.9). With

respect to religiosity, a majority of participants reported being atleast somewhat religious or spiritual (n = 19) and the remainderreported no religiosity (n = 9).6 US participants also reported theirchildhood places of residency, either in an urban setting (a city ortown) (n = 11) or a non-urban setting (a suburb, small town, or coun-tryside) (n = 17). Participants completed informed consent andreceived partial course credit for participating.

4.3. Procedure

In Panama, participants were interviewed in Ngöbere or Span-ish by two authors (S.G.G. and b.l.o). In the US, participants com-pleted a computer-based survey in English or Spanish. Thesewere selected as the most culturally appropriate methods for eachcontext, reflecting typical methods used in small-scale communi-ties and university settings, respectively.

The Spanish protocols were independently forward- and back-translated between English and Spanish by two trained researchassistants in the US; before use in Panama, protocols werereviewed for local accuracy by S.G.G. Ngöbere protocol translationwas also conducted by S.G.G., who has linguistic expertise in

6 Demographic data were available from 28 of the 34 US participants.


Ngöbere. A second community member trained in research meth-ods reviewed and back-translated critical phrases of the Ngöbereprotocol. During Ngöbere interviews in Panama, S.G.G. translatedthe participant’s responses immediately into Spanish. Interviewresponses were audio-recorded and later transcribed and trans-lated from Spanish into English by a trained bilingual researchassistant in the US.

7 Among our Ngöbe sample, differences were found across the two plant stimulusversions (original and culture-specific). Participants were more likely to predict goal-directed action for the culturally familiar ground vine (M = 0.80, SD = 0.42) than theoriginal indoor potted plant (M = 0.31, SD = 0.48) across both BP and BA conditions, F(1,19) = 5.16, p < 0.05, g2 = 0.21. However, this effect held for all three plant itemsalthough the indoor/outdoor plant was the only item that differed across versions, sothe observed difference cannot be due to the familiar versus unfamiliar item. Thisfinding may instead reflect participant demographics, given that participants in thefamiliar plant stimulus condition also tended to have less schooling (correlated withstimulus version r = �0.62, p < 0.01) and were also more likely to be members of theNgöbe church (r = 0.59, p < 0.01).

8 Among US participants, females predicted more goal-directed actions formachines (M = 0.56, SD = 0.41) than males (M = 0.19, SD = 0.24) in the BP conditiononly, F(3,87) = 2.97, p < 0.05, g2 = 0.09. This gender difference was unexpected.Action predictions for the other three kinds did not differ.

9 One additional US participant did complete Exp. 1A but was excluded from Exp.1B due to experimenter error.

5. Experiment 1A: Goal-directed action predictions

Acting toward self-beneficial goals represents a basic survivalbehavior that is a pervasive, readily observable characteristic of liv-ing kinds (Opfer, 2002; Opfer & Gelman, 2001). As such, plant goal-directed behavior should be recognized by both US and Ngöbeinformants even if, as we hypothesize in Experiment 1B, it willbe interpreted under different frameworks. Therefore, we pre-dicted cultural agreement on action predictions for plant goal-directed agency in Experiment 1A.

5.1. Participants

Participants were 23 Ngöbe (7 females) and 33 US participants(16 females). One additional US participant was excluded due toexperimenter error. Gender was not found to have an effect onthe primary findings for any experiments. Where gender effectswere found for secondary analyses, those are reported in footnotes.

5.2. Materials and methods

Stimuli were colored line drawings of plants, animals, machi-nes, and simple artifacts (three of each kind) from Opfer andGelman (2001) (see original study for materials). Participants pre-dicted whether each entity would change its current direction ofmovement in response to the presence of an object, representingeither a self-beneficial goal object such as food or sunlight (BenefitPresent condition, BP) or a neutral object, such as a rock or picture(Benefit Absent condition, BA). In the BP condition, predicting thatthe entity would change its direction of movement towards thebeneficial object was coded as a goal-directed action prediction(scored as 1). Otherwise, predicting that the entity would notchange its direction of movement (toward a neutral object in theBA condition, or at all in the BP condition) was coded as no goal-directed action (scored as 0). Following the original study, the pur-pose of the BA condition was to provide a control where changes inmovement would not be systematically predicted, to strengthenthe interpretation that predictions for a change in movement forthe BP condition are genuinely goal-directed judgments. Stimuliwere randomly presented within randomized (US) or counterbal-anced (Panama) kind blocks.

Culturally appropriate adjustments were made to select stimuliin Panama to avoid potential confounds related to item familiarity(see Appendix A.1). A few modest changes were made part waythrough the study. We began our interviews with the originalindoor potted plant item (N = 13) but then replaced it with an out-door ground vine (conserving the growth movement pattern andgoal of sunlight) for subsequent interviews (N = 10). Ngöbe donot grow plants indoors, so the vine represented a more familiarcontext for reasoning about sunlight-directed growth patterns.Where differences are found for the plant item, analyses arereported in footnotes. We also replaced the original novel machineitems with three familiar machines to ensure that Ngöbe infor-mants were not reasoning based on inferred special properties ofnovel machines. There were no differences in action predictionsor agency attributions for novel (N = 10) versus familiar (N = 13)machine stimuli.

5.3. Results

We predicted cultural agreement on predictions for goal-directed action, such that plants and animals, but not machinesand artifacts, will be viewed as capable of acting toward self-beneficial goals. There should also be an overall agreement thatagents will not systematically act toward neutral objects, as thisdoes not represent a case of genuinely goal-directed behavior.Goal-directed action predictions were averaged across the threeentities for each kind. Tests were conducted with a 4-factor (kind:animal, plant, machine, other artifact) repeated-measures ANOVAon action predictions, with condition (benefit present versusabsent; BP, BA) and culture as between-subjects variables. Interac-tions were investigated with separate analyses for either kind orcondition, as appropriate. Follow-up tests were conducted withBonferroni corrections.

As expected, there was broad cultural agreement on action pre-dictions. No reliable cultural differences emerged for any individ-ual item or kind (all ps > 0.10) (see Table 1).7 Both Ngöbe and USparticipants expected plants and animals to act toward goals moreoften than machines and simple artifacts, and to do so more whena benefit was present than when it was not. The main effect of kind,F(3,156) = 97.42, p < 0.01, g2 = 0.65, when broken down by specificcontrasts, indicated that participants predicted goal-directed actionsfor animals more than all other kinds (M = 0.85, SD = 0.27), and alsopredicted more goal-directed action for plants (M = 0.61, SD = 0.38)than machines (M = 0.20, SD = 0.32) or artifacts (M = 0.17,SD = 0.25) (all p’s < 0.05) (machines and simple artifacts did not dif-fer).8 Additionally, participants were more likely to predict goal-directed actions in the BP condition (M = 0.59, SD = 0.16), than theBA condition (M = 0.32, SD = 0.16), F(1,52) = 39.67, p < 0.01,g2 = 0.43, indicating cross-cultural agreement that self-beneficial(as compared to neutral) objects prompt more goal-directed actionsamong both plants and animals.

5.4. Discussion

We expected and found cross-cultural agreement on predic-tions concerning the kinds of agents that can act toward self-beneficial goals—namely, plants and animals, but not machinesor artifacts. We now turn to the critical question of how suchaction capacities are interpreted.

6. Experiment 1B: capacity attribution

After completing Experiment 1A, the same participants judgedfive agency capacities (see, want, feel, think, feel pain) of eachentity (following Opfer & Gelman, 2001).9 The key question con-cerns capacity attributions to plants. We predicted that Ngöbe butnot US participants would extend capacity attributions (e.g., think-

0.4

0.6

0.8

1

ean

a�ri

butio

ns

NgöbeUS

Table 1Predictions of goal-directed action by culture, kind, and condition.

Condition/Culture Animals Plants Machines Artifacts

Mean SD Mean SD Mean SD Mean SD

Benefit-Present ConditionNgöbe 1 0 0.82 0.35 0.18 0.40 0.33 0.30US 0.96 0.11 0.91 0.25 0.37 0.38 0.11 0.26

Benefit-Absent ConditionNgöbe 0.72 0.4 0.36 0.22 0.14 0.22 0.17 0.22US 0.69 0.27 0.29 0.21 0.06 0.13 0.1 0.16


ing, feeling pain) to plants as well as animals. By hypothesis, this dif-ference arises because these predicates correspond to two differentsets of meanings (communicative or psychological) on the two cul-tural frameworks for agency. If agency capacities are conceptualizedunder a folkcommunicative framework focused on interactions withthe environment, then capacity attribution profiles should keepplants closer to animals than to inanimate machines and simple arti-facts (Ngöbe). If these capacities are conceptualized under a folkpsy-chological framework focused on having a mind, then plants shouldbe closer to other inanimates than they are to animates (US).

0

0.2

Animals Plants Machines Artifacts

M

Kind category

Fig. 1. Capacity attributions by kind and culture.


Stimuli were the same as in Experiment 1A (including theculture-specific plant and machine items). Participants were askedif each entity could see, want, feel, think, and feel pain (see Appen-dix A.2). Items were blocked by kind in counterbalanced (Panama)or randomized order (US) and capacity probes were presented incounterbalanced order.

6.2. Results

Capacity attributions were averaged across the three entities bykind. To explore cultural differences, attributions were analyzedwith a 2 (culture) by 4 (kind) by 5 (capacity) repeated-measuresANOVA. Here we report only the focal cultural contrasts; for fullresults, see Appendix A.3.

As shown in Fig. 1, the critical interaction of culture by kind wassignificant, F(3,162) = 13.09, p < 0.001, g2 = 0.20. Follow-up com-parisons (Bonferroni-corrected) showed that Ngöbe attributedmore capacities to plants (M = 0.61, SD = 0.33) than did US partici-pants (M = 0.26, SD = 0.20), t(32.75) = 4.62, p < 0.001, d = 1.61. Cul-tural patterns of attribution to artifacts also differed, t(22.30)= 2.32, p < 0.05, d = 0.98, but not to animals or machines (seeAppendix A.3).

As is evident from the overall attribution profiles in Fig. 1, bothgroups differentiated among animals, plants, and artifacts (machi-nes and simple objects), but the distinctions drawn between thosecategories fell at different points. ForNgöbe, plants possess substan-tial agency capacities and stand closer to animals than to inanimateartifacts. By contrast, US participants denied almost all capacities toplants, placing them closer to inanimate artifacts than animals.

On the proposed folkcommunication framework, one predictionis that mappings between action predictions and capacity attribu-tions should be nearly one-to-one because the ability to interactmay itself be interpreted as communication. If so, we should findthat goal-directed action predictions correlate with agency attribu-tions among Ngöbe participants. This was in fact the case for allthree non-animal kinds: action predictions and capacity attribu-tions were correlated for plants (partial r = 0.47, p < 0.05, df = 20,controlling for BP/BA condition), artifacts (partial r = 0.49,p < 0.05. df = 20) and machines (partial r = 0.43, p < 0.05, df = 20).In contrast, US individuals should not evidence this same mapping

because action is selectively linked to folkpsychological agencyattributions, as action is interpreted differently for different kindsaccording to domain-specific frameworks (physical, biological, ormental). Consistent with this, there were no reliable correlationsbetween action predictions and agency attributions for any kindamong US participants. These observations indicate that the map-ping from goal-directed action to agency attribution is more directfor Ngöbe than for US participants, in line with the idea that Ngöbeinfer agency on the basis of a capacity to interact.

Ngöbe informants also made more capacity attributions acrossall four kinds than US participants, F(1,54) = 15.26, p < 0.001,g2 = 0.22, and this trend held reliably for every capacity exceptthinking, ps < 0.05. The three-way interaction among culture,capacity, and kind, F(12,648) = 4.59, p < 0.001, g2 = 0.08, indicatedthat cultural attributions for each kind varied according to specificcapacity. To illuminate these patterns, a 2 (culture) by 5 (capacity)repeated-measures ANOVA was performed on attributions for eachkind separately; here we report findings for the key test case ofplants (see Appendix A.3 for other kinds).

Plants. As indicated in Fig. 2, Ngöbe were more likely than USparticipants to attribute every capacity except thinking to plants.Ngöbe endorsed plant capacities for feeling, wanting, and pain atlevels well above chance (2-tailed ps < 0.05), while thinking wasendorsed slightly above chance (ns). In contrast, US attributionsdropped off precipitously for every agency capacity except feeling,with wanting endorsed at low levels (ns), and pain and thinkingstrongly rejected (2-tailed ps < 0.01). Notably, Ngöbe were overseven times more likely than US participants to say that plantscan experience pain.

6.3. Discussion

In Experiment 1B, our key prediction concerning plant agencywas borne out: Ngöbe were far more likely than US participants

0

0.2

0.4

0.6

0.8

1

Feel Want Pain Think See

Mea

n a�

ribu

tions

Capacity

Ngöbe

US

Fig. 2. Capacity attributions to plants by culture.


to attribute agency capacities to plants. Not only was this culturalpattern robust across every capacity except thinking (which wasstill a strong trend), but also the effects were large enough to sug-gest qualitatively distinct models for plant agency. On the Ngöbemodel of agency attribution, plants stand in closer relation to ani-mate agents than they do to inanimate artifacts. In direct contrast,the US model denies most capacities to plants, placing them in clo-ser relation to inanimate objects than to animals.

Interestingly, most participants in both groups agreed thatplants can feel, replicating Opfer and Gelman’s findings with UScollege students. It is unclear to what extent this represents aquasi-psychological construal of plants for US participants, or ‘‘ametaphor in the biological domain” as Opfer and Gelman suggest(p. 1380). Likewise, both US and Ngöbe participants tended toagree that plants cannot see, evidently interpreting sight as a visualcapacity requiring eyes. Presumably, attributions would have beenhigher for a more neutral probe such as ‘‘detect” or ‘‘perceive.” Itmay be that the observable relationship between eyes and sightsupports cross-culturally shared intuitions about this capacity(see also Meltzoff & Gopnik, 2013). By contrast, the properties(e.g., mind) underlying folkpsychological construals of capacitiessuch as wanting or pain are less observable, and this may con-tribute to greater cross-cultural divergence in corresponding intu-itions. We speculate that ‘‘thinking” may have elicited relativelyfewer attributions among Ngöbe because it rarely lends itself toexpression through an observable relationship such as wanting(e.g., growing toward sunlight) or pain (e.g., producing sap whencut), so could be less relevant to a communicative folk theory.

A limited number of Ngöbe informants also endorsed certaincapacities for artifacts and machines, producing small but reliablecultural effects in agency attributions for those two kinds. Thoughthese attributions may suggest a positive response bias for someNgöbe participants, the effects were not of sufficient magnitudeto account for the focal cultural differences in plant agencyattributions.

10 We describe these plant interactions as communication, kin recognition, and kinaltruism following the scientific botanical literature (e.g., Bhatt et al., 2011; Biedrzycki& Bais, 2010; Falik, Mordoch, Quansah, Fait, & Novoplansky, 2011), while recognizingthat some readers may prefer leaner terms such as signaling.

6.4. Experiment 1: overall discussion

Experiment 1 showed that Ngöbe and US participants both rec-ognize goal-directed action among plants and animals, but theinterpretations given to this agency differ across cultures. On theUS model, these agency capacities are seen as unique to animalsso that plant goal-directed action does not warrant attribution ofpsychological states. The Ngöbe results, by contrast, indicate thatthese agency capacities are construed as properties shared to a

substantial extent by both plants and animals. By hypothesis, thisreflects an alternative interpretive framework: Ngöbe view plantgoal-directed action as evidence of a relational capacity for com-munication. Ngöbe capacity attributions represent a departurefrom the domain-specific view that plants are denied the inten-tional capacities typically associated with animals. In Experiment2 we tested the idea that different frameworks for agency willafford distinct inferences and interpretations in a novel domainof plant behavior—social agency.

7. Experiment 2A: action predictions for plant socialinteractions

Experiment 2 focuses on concepts of plant social agency, specif-ically plant communication and kin altruism.10 These social capaci-ties were selected because they represent a recent area of Westernbotanical science where the interpretation of findings is still subjectto debate. Not surprisingly given the recentness of these discoveries,plant social interactions are subtle and less directly observable thanthe self-beneficial goal-directed actions from Experiment 1. Presum-ably then, plant social interactions arenot commonknowledge amonglaypersons. This provides a test bed for assessing how people reasonabout plant behavior on the basis of intuitive conceptual frameworksrather than simple fact retrieval. We hypothesize that Ngöbe shouldbe more likely than US participants to predict that plants will engagein social behaviors, on the basis of an ecocentric folktheory that pro-vides a lens for observing and expecting complex communicativeexchanges among organisms and their environments.

As in Experiment 1, illustrated scenarios were used to assessaction predictions (Experiment 2A) and corresponding capacityattributions (Experiment 2B) concerning communication and kinaltruism. Unlike Experiment 1, however, we now expected actionpredictions to differ in addition to capacity attributions.

7.1. Participants

Participants were 17 Ngöbe (5 female) and 30 US (13 female)participants from Experiment 1. Ngöbe samples were smaller thanin Experiment 1 because pilot interviews were conducted to refinethe protocol (see Appendix C.1).


Participants were presented with four line drawn sequences ofplants engaged in events characteristic of communicative signalingand kin-selective resource sharing (see Fig. 3; complete materialsin Appendix C.2). For each scenario, participants were told thatthe plants were connected via roots and asked to predict whetheror not one plant would modify its activity in response to anotherplant. Thus they always chose between two outcomes: either theplants would interact, leading to an adaptive change for one plant,typically the receiver plant (coded as a social interaction predic-tion, scored as 1); or there would be no interaction between theplants and therefore no adaptive change for either plant (codedas no social interaction prediction, scored as 0).

The two communication scenarios involved root-connectedplants signaling in response to drought or an insect plague,prompting an anticipatory adaptive response in the receiving plant(modeled on experimental results reported in Falik et al., 2011;Heil & Karban, 2010) (see Fig. 3). In each scenario, participants pre-dicted whether or not the receiver plant(s) would already be in a

Fig. 3. Illustration of prediction scenario for plant interactions.

Fig. 4. Action predictions for plant social interactions by culture.


prepared state prior to the arrival of the threatening event. The twokin altruism scenarios involved resource sharing, either betweentwo sibling plants or a mother tree and her seedling (modeled onexperimental results and observations reported in Dudley & File,2007; Simard, 2009; Wilkinson, 1998). In each scenario, partici-pants predicted whether or not the more mature plant would mod-ify its resource consumption such that the less mature plant wouldbenefit. For all four scenarios, the experimental logic was thataction predictions would be made on the basis of inferred commu-nication or kin resource sharing, respectively. Analysis of explana-tions confirmed that most participants did follow this reasoning(see Appendix C.3).11

In Panama, prediction scenarios were blocked by domain, withkin altruism (mother tree, then sibling plants) appearing beforecommunication (insects, then drought) for half the participants.In the US, the communication block was always presented first(insects, then drought) followed by the kin altruism block (siblingplants, then mother tree).

7.3. Results

As indicated in Fig. 4, Ngöbe were more likely than US partici-pants to predict that plants would communicate and shareresources with kin. The two prediction scenarios in each domain(communication, kin altruism) were conceptually similar andexhibited similar results (see Appendix C.3), so a composite mea-sure was created by averaging across those two predictions foreach participant. Instances where participants did not offer a pre-diction (‘‘don’t know” or no opinion) were treated as missing data(2 US participants, 1 Ngöbe participant).

Communication: Ngöbe were reliably more likely to predict thatconnected plants would be prepared for the arrival of drought or

11 For the drought communication scenario, some participants explained that theconnected plant would survive the drought due to water sharing between plants, thusinferring that the interaction involved resource sharing rather than communication.These responses were analyzed in the same manner as other communicationpredictions as they are consistent with an overall focus on plant social behavior.

insects (M = 0.91, SD = 0.26) than were US participants (M = 0.70,SD = 0.31), F(1,45) = 5.59, p < 0.05, g2 = 0.11.12 Kin altruism: Ngöbewere reliably more likely to predict that plants would shareresources with kin (M = 0.59, SD = 0.32) than US participants(M = 0.32, SD = 0.40), F(1,45) = 5.66, p < 0.05, g2 = 0.11.

The finding that a majority of both Ngöbe and US participantspredicted plant communication raises the question of how suchinteractions among plants are interpreted. To explore this hypoth-esis, we analyzed explanations for positive communication predic-tions (excluding participants who predicted no communication).We predicted that Ngöbe participants would explain plant commu-nication as a genuinely social interaction involving intentionaltransactions directed toward other plants, whereas US participantswould be more likely to see communication as a byproduct of indi-vidual survival goals or reductive biological mechanisms.

12 Gender effects were found for kin altruism predictions among US participants, F(1,28) = 4.12, p = 0.05, g2 = 0.13, such that males (M = 0.44, SD = 0.43) were morelikely than females to predict kin altruism (M = 0.15, SD = 0.32). There were noreliable effects of gender among Ngöbe participants.


Each explanation received one of three mutually exclusivecodes: (1) social goal, (2) individual goal, or (3) (non-goal-directed) biological and/or physical mechanisms (see coding tablein Appendix C.4). Explanations that identified a purposive action ofone or both individual plant(s) and linked this to a beneficial out-come (e.g., adaptation, survival, growth) were coded as goal-directed. Goal-directed explanations then could fall into one oftwo categories. An explanation that cited an other-beneficial goalas the cause or purpose of one plant’s interaction with anotherplant was coded as a social goal (e.g., signaling for the receivingplant’s benefit; helping or protecting another plant; such as, ‘‘Theywere communicating underground and telling the other tree, theywere warning it that it has to save some water.”). An explanationthat cited self-beneficial goals as the cause of plant interactionwas coded as an individual goal. In such cases, the beneficial out-come was facilitated by plant interaction (e.g., root contact), butit was a byproduct rather than the goal of that interaction (e.g.,‘‘They sense the chemicals and it’s a natural response to survive”).Explanations that did not identify any goal but instead focused onnon-purposive causes of action (biological processes, physics, orgeneric factors) were coded as mechanistic (e.g., innate reactions,automatic processes, or physical mechanisms, as well as species-level evolutionary mechanisms).

As indicated in Table 2, there were reliable cultural differencesin interpretations of plant communication for both scenarios(Drought: v2 (1, N = 36) = 9.24, exact, 2-sided p < 0.01; Insects: v2

(1, N = 37) = 18.60, exact, 2-sided p < 0.01).13 Ngöbe were far morelikely to explain plant communication in terms of social goals(20/31) than individual goals (8/31) or non-intentional mechanisms(3/31). By contrast, US participants rarely explained plant communi-cation as a social interaction (2/42) and instead favored individualgoals (20/42), or non-intentional mechanisms (20/42). US respon-dents in particular often stressed the non-intentional nature of plantactivity, as in, ‘‘Some sort of chain reaction thing. . .no thinking, justreactions.”

Summary. Experiment 2A revealed cultural differences in infer-ences about plant social behavior. Ngöbe were more likely than USparticipants to predict both plant communication and kin altruism.Furthermore, analysis of explanations showed that Ngöbe weremore likely to interpret plant communication in terms of socialgoals, whereas US participants tended to interpret communicationas a byproduct of individual survival goals and biological mecha-nisms. In the next experiment, these cultural beliefs were furtherinvestigated with direct probes for capacity attributions concern-ing plant social agency.

8. Experiment 2B: capacity attributions for plant socialinteractions

Experiment 2B involved judgments of four agency capacitieslinked to plant communication (signaling) and kin altruism (kinrecognition, helping, feeling another plant’s pain). We predictedthat Ngöbe would be more likely than US participants to attributethese social agency capacities to plants, particularly those associ-ated with kin altruism, in keeping with their greater recognitionof complex social behaviors among plants.

8.1. Participants

15 Ngöbe (5 female) and 30 US (13 female) participants fromExperiment 2A also participated in Experiment 2B.

13 We report the N � 1 chi-square test rather than Pearson’s chi-square, because theN � 1 is more robust to low expected counts and is recommended for small samples(Campbell, 2007). The linear-by-linear association test generated in SPSS CROSSTABSis equivalent to the N � 1 chi-square (Weaver, 2013).


Participants judged whether plants are capable of four capaci-ties: (1) recognizing kin, (2) helping offspring via roots, (3) signal-ing or communicating, and (4) feeling pain for another injuredplant (see full probes in Appendix C.4). Each item was presentedimmediately after the matching illustrated prediction scenariofrom Experiment 2A (and in the same order) such that Experiments2A and 2B were interwoven. The pain item was always presentedlast, accompanied by a photograph of a fallen tree surrounded byother trees.

8.3. Results

Capacity attributions were scored as yes or no, excluding ‘‘don’tknow” responses. The data from one Ngöbe participant who pro-vided 2 such responses were excluded from analysis. Attributionswere analyzed with a 2 (culture) by 4 (capacity) repeated-measures ANOVA. Only the focal cultural contrasts are reportedhere; for full results, see Appendix C.5.

As indicated in Fig. 5, Ngöbe were more likely than US partici-pants to attribute social agency capacities to plants (NgöbeM = 0.84, SD = 0.27; US M = 0.38, SD = 0.27), F(1,42) = 27.39,p < 0.001, g2 = 0.40. This was reliable for each capacity individually(Fs (1,42) = from 6.67 to 68.28, all ps < 0.05), except plant signaling,where the cultural trend was only marginally significant, F(1,42)= 2.90, p = 0.10, g2 = 0.07.

The interaction of culture by capacity, F(3,126) = 3.47, p < 0.05,g2 = 0.08, showed that the extent of divergence between US andNgöbe attributions differed across particular capacities, with indi-vidual social capacity attributions varying more among US thanNgöbe participants. US attributions did not extend to feelinganother plant’s pain, leading to reliable interaction effects, F(3,87) = 16.12, p < 0.001, g2 = 0.36. In contrast, Ngöbe attributionswere uniformly high (all were above chance, ps < 0.05, except kinrecognition p = 0.07), and there were no reliable differences inattributions across the four capacities, F(3,39) = 0.07, ns (for pair-wise comparisons, all ps > 0.99).

These results provide converging evidence for the key findingsfrom Experiment 2A. Most US and Ngöbe participants agree thatplants are capable of communicative signaling, but only Ngöbeare likely to agree that plants also possess capacities linked tokin altruism, including kin recognition, resource sharing, and feel-ing another plant’s pain.

8.4. Experiment 2: discussion

In Experiment 2, we found converging evidence from actionpredictions and capacity attributions that Ngöbe are more likelythan US individuals to recognize complex social behaviors ofplants. In line with recent botanical science, the majority of Ngöbeinformants predicted that plants would communicate and shareresources with kin. These findings extend the observed cultural dif-ferences in capacity attribution from Experiments 1B and 2B bysuggesting that different ways of interpreting plant agency carrycognitive consequences for inferential reasoning about actual plantbehavior.

9. Language

We conducted exploratory analyses of language differences inthe capacity attribution tasks in both Experiments 1B and 2B toassess whether these different predicate assignment patternsreflect cross-linguistic differences among Spanish, English, andNgöbere. If attribution patterns hold across languages within each

Table 2Interpretations of plant communication for each prediction scenario.

Scenario/Culture Social goal Individual goal Mechanistic Total

Drought scenario (N = 36)a

Ngöbe 9 (60%) 3 (20%) 3 (20%) 15US 1 (5%) 10 (48%) 10 (48%) 21

Insect scenario (N = 37)a

Ngöbe 11 (69%) 5 (31%) 0 (0%) 16US 1 (5%) 10 (48%) 10 (48%) 21

a N = number of participants who made a positive action prediction for that scenario (excluding explanations given by participants who predicted no action).

0

0.2

0.4

0.6

0.8

1

Signaling Kin recognition Mother tree helping

Feel other's pain

Mea

n a�

ribu

tions

Capacity

Ngöbe

US

Fig. 5. Social capacity attributions to plants by culture.


of our cultural samples, then this would offer preliminary evidencethat cultural differences in capacity attribution do not merelyreflect linguistic artifacts.14

The US sample included 16 bilingual English-Spanish speakers,of whom 12 completed the study in Spanish and 4 in English (seeAppendix B). An additional 18 English speakers rounded out our USEnglish sample (for n = 22). In Panama, all participants spoke Span-ish and Ngöbere (most with bilingual fluency levels), so approxi-mately half completed each set of tasks in Ngöbere (n = 13 inExp. 1; n = 11 in Exp. 2) and the other half in Spanish (n = 10 inExp. 1; n = 6 in Exp. 2). Language assignment was sometimes basedon considerations of language proficiency; otherwise participantswere randomly assigned to a language condition. Language couldnot be randomly assigned in the US due to the small number ofbilingual participants.

Results. There were no reliable language effects of Spanish ver-sus English among US participants, or for Spanish versus Ngöbereamong Ngöbe participants. This indicates that the attribution ofcapacities such as wanting, feeling, or communicating to plants isunlikely to be a linguistic artifact. These results must be inter-preted with caution given our modest sample sizes and nonran-domized language assignment. Nonetheless, they providepreliminary evidence that agency attributions are not solelydependent on language for either Ngöbe or US participants.

14 We did not expect that language would affect the action prediction tasks(Experiments 1A and 2A), which rely on predictions of one or another outcome ratherthan interpretations of particular predicates. Action scenarios intentionally usedneutral event descriptions that avoided the critical agency terms in question (i.e., noreferences to want, kin altruism, communicate, etc.).

10. Detecting multiple conceptual models

In this section we consolidate our findings from the aboveexperiments by using cultural consensus modeling to address sev-eral critical questions. One outstanding question concerns the nat-ure of conceptual divergence between US and Ngöbe views ofagency. One could argue that the cultural mean differences inagency attribution do not decisively demonstrate two distinct con-ceptual models. For instance, Ngöbe attributions for the variousagency capacities in Experiment 1B are not consistently near ceil-ing for plants as they are for animals, raising the possibility thatNgöbe respondents share the basic animal-oriented model of USparticipants but are simply more likely to endorse scattered capac-ities for plants. In other words, the mean differences between USand Ngöbe responses could represent variation from a single (uni-versal) model of agency that is animal-oriented, rather than point-ing to a distinct model where agency is equally a property of plantsand animals. Another important question concerns the sources ofindividual variability that contribute to these differences bothwithin and across cultures.

To answer these and relatedquestions,weused a newbottom-upcultural consensus modeling method to identify whether multipleconceptualmodels of agencyare latent inparticipant responses (col-lapsing across US and Ngöbe participants), and to explore between-and within-culture variations on these conceptual models. Ourmodeling is conducted on data from both capacity attributions andaction predictions across the two experiments for a comprehensiveassessment of agency frameworks. Our first objective was to deter-mine if there are one or more consensus conceptual models presentamong our respondents. This is an important step beyond the cul-tural differences reported above because two groups can differ evenwhile one group fails to converge on a consensus model (Le Guen,Iliev, Lois, Atran, & Medin, 2013). If more than one consensus isdetected thiswill indicate thatmultiplemodels are present. Second,we are interested in the agency attribution profile of each consensusmodel,which can answer the key question ofwhether or not there isan alternativemodel that consistently endorses plant agency on parwith animals. Third, we are interested in how participants clusterinto the consensus models according to their actual cultural mem-bership and individual characteristics, as well as how individual-level variablesmaypredict competencewithin eachmodel. Thiswillenrich understanding of across- and within-group variability andpoint to individual factors that may influence one’s conceptualmodel of plant agency.

10.1. Cultural consensus modeling

Cultural consensus modeling (CCM) (Romney, Weller, &Batchelder, 1986) has been applied extensively in prior culturalresearch on biological cognition (Atran & Medin, 2008; Medin,Ross, Cox, & Atran, 2007). Conceptually similar to factor analysis,CCM determines if a single underlying model exists, as well aswhether patterns of residual agreement beyond the overall con-sensus suggest additional models that hold for subgroups in the


sample. Each participant is assigned a cultural ‘‘competence,”where higher scores on a factor loading indicate that an individ-ual’s responses are closer to the cultural consensus. CCM also cal-culates the consensus model ‘‘answer key” with a competencyweighted consensus rating for each item, providing a more reliableapproximation of the common truth than traditional data aggrega-tion techniques (France & Batchelder, 2015).

Recent advances in CCM provide a model-based way to derivemulticultural consensus from continuous response data (Anders,Oravecz, & Batchelder, 2014). This new Bayesian inferentialapproach to consensus modeling (Anders, 2013) offers a keyadvantage over prior CCM techniques by treating culture as alatent variable and using a bottom-up approach to identifying dis-tinct consensus models and their members (rather than relying onknown cultural membership as in traditional CCM methods). Thisconfers a theoretical advantage over statistical models that treatculture as a fixed independent variable, by focusing on conceptualmodels and allowing cultural membership to remain latent. Toemphasize this analytic distinction, we refer henceforth to ‘‘con-ceptual/consensus models” rather than ‘‘cultural models.” CCTpackalso introduces a new response precision parameter to assesswhether some items are more difficult than others, unlike tradi-tional CCM methods that assume all items are equally difficult.

10.2. Methods and procedure

Using the CCTpack R software package (Anders, 2013), weapplied the Bayesian CCT model to the combined capacity attribu-tions from Experiments 1 (5 individual capacities for each of 4kinds) and 2 (4 plant social capacities) and action predictions fromExperiment 2 only (4 plant social interaction predictions) (totalk = 28). Experiment 1 action predictions were not included becausethey were confounded by experimental condition. The participant-level response data from all US and Ngöbe participants together(N = 47) were subjected to cultural consensus modeling for a 47(participants) � 28 (items) matrix. For clarity, results are presentedat the item level for plants but the kind level for the other nonhu-man kinds (as in Fig. 6).

In CCTpack, two posterior predictive checks are used to com-pare which model best fits the actual data. 1. To check that the con-sensus structure of the data is appropriately fit by the model (i.e.,the appropriate number of consensuses), a scree plot of eigenval-ues assesses the fit between the model-generated estimates andthe actual structure of the data (see Anders, 2013 for details). 2.To check whether the item difficulty parameter should be treatedas uniform (i.e., homogenous item difficulty) or variable (i.e.,heterogeneous item difficulty), a variance dispersion index (VDI)reports how well the model captures differences across itemsdue to response variability; these VDIs are compared and themodel with better fit is selected (see Anders et al., 2014).

16 A 3-consensus model was unable to reach convergence of modeling chains(Anders, 2014), indicating that there were no detectable third components and/or toomany degrees of freedom.17

10.3. Results

10.3.1. Model selection and identification of latent conceptual modelsThe data were simulated using a series of models with 1, 2, and

3 consensuses, with and without heterogeneous item difficulty.Item difficulty. The VDI check for item difficulty indicated that a

model assuming homogeneous item difficulty fit the data, whereasa model with heterogeneous item difficulty failed to fit the data.15

Consensus fit. The scree plots of eigenvalues showed that thefirst factor was substantially larger than subsequent factors, witha second factor evident but less pronounced. Posterior predictive

15 A model that assumed heterogeneous items difficulty was unable to reachconvergence of modeling chains for a successful fit to the data (Anders, 2014).

checks showed that the data structure was well fit by both aone- and two-consensus model, but a three-consensus modelfailed to fit the data.16 This suggests an overarching consensusaccompanied by two subgroups. The overall consensus was expecteddue to the substantial agreement on capacity attributions for ani-mals, machines, and artifacts. A two-consensus model would beexpected to pick up additional variability concerning plant agency.In the past, we have consistently found a good overall consensuspaired with systematic subgroup differences (Medin et al., 2007).Thus, we selected a two-consensus model with homogeneous itemdifficulty.

10.3.2. Consensus modelsThe model identified two consensuses latent in the agency rank

responses (see Fig. 6). With respect to the question of whetherthere are two different models for plant agency, the results areunambiguous. The first consensus model presents a strong andconsistent endorsement of the nine plant capacities for social andintentional agency. Indeed, with the exception of seeing, capacityattributions to plants and animals are indistinguishable (at or nearceiling). With respect to action predictions, inferences for kin altru-ism were lower than would be expected given that the associatedcapacities (kin recognition, mother tree helping) are stronglyendorsed. This indicates that the kin altruism prediction scenariosmay have tapped other considerations for these participants (e.g.,resource sharing weighed against access to sunlight in the mothertree scenario). The second consensus model presents a highlyselective endorsement of plant agency. Beyond a capacity to signaland feel, plants are not viewed as having much agency and areclearly less agentic than animals.

10.3.3. Consensus model membershipConsensus group membership. The model detected and clustered

individual participants into the two consensus models in close cor-respondence with actual cultural membership. Members of theplant-oriented Model 1 were primarily Ngöbe respondents (9 of11) whereas members of the animal-oriented Model 2 were pri-marily US respondents (28 of 36). The remaining members of eachmodel were from the other cultural group.

Individual differences in consensus model membership. Ngöbe. TheNgöbe sample exhibited substantial clustering variability (9 clus-tered with Model 1, 8 with Model 2), which may speak to the pres-ence of multiple models of agency that correlate with particularindividual characteristics. Based on prior research, we anticipatedthat Ngöbe members of the plant-oriented Model 1 would leadmore traditional community-based lifestyles than Model 2 mem-bers, as indexed by three demographic measures (Indigenouschurch membership, age, formal schooling). Consistent with ourprediction, members of Model 1 were more likely to be membersof the Indigenous church, whereas members of Model 2 were morelikely to be Evangelical churchgoers or non-attendees, F(1,15)= 5.79, p = 0.05, g2 = 0.28. Model 1 members also tended to havefewer years of formal schooling, F(1,15) = 3.33, p = 0.09, g2 = 0.18,and were more likely to be male, F(1,15) = 3.33, p = 0.09,g2 = 0.18, but these trends only approached significance.17

In addition to individual characteristics, we also checked to seewhether any task factors were associated with model membership.Importantly, task language (Ngöbere or Spanish) did not differenti-

Note that age did not correlate with model membership among Ngöbe, suggestingthat age is not a primary factor in these conceptual differences among our Ngöbesample. This is a useful observation in light of any potential concerns regarding theage differences between our Ngöbe and US samples.

.00

.20

.40

.60

.80

1.00

Com

mun

icat

ion

Kin

Altr

uism

Feel

oth

er's

pain

Kin

Rec

ogni

tion

Mot

her t

ree h

elpi

ng

Sign

alin

g

See

Feel

Pain

Thin

k

Wan

t

Ani

mal

agen

cy

Mac

hine

age

ncy

Art

ifact

age

ncy

Action prediction

Capacity a�ribution Capacity a�ribution

Plants Other nonhumans

Fig. 6. Consensus models for agency of plants and other nonhumans.


ate between members of Model 1 and 2. This provides further evi-dence that these conceptual models of agency are unlikely toreflect linguistic differences between Spanish and Ngöbere. How-ever, Model 1 members were more likely to be in the culture-specific stimulus version of Experiment 1, F(1,15) = 6.48, p < 0.05,g2 = 0.31. As noted earlier, assignment to this version was inciden-tally correlated with other demographic factors, so the significanceof this finding is unclear.

US. US participants clustered heavily into a single model (28 of30 with Model 2), which limited analysis of differences betweenModel 1 versus Model 2 members due to the small sample (Model1, n = 2). Unsurprisingly, we found no reliable differences in modelmembership by either individual characteristics or task language.On the other hand, it is a notable finding that US members con-verged so strongly on a single model of agency despite being thelarger sample and being drawn from two distinct languagesamples.

Conceptual differences in cross-model membership. Given that acultural majority emerged for each consensus model, we wereinterested in the responses of those individuals who were cross-clustered with the other consensus model (i.e., that of their non-majority culture). The agency attribution profiles of these sub-groups could speak to what happens at the boundaries of each con-ceptual model. All differences reported below were significant atthe p < 0.05 level in MANOVA’s comparing agency attribution mea-sures across the two models separately for each culture.

Ngöbe participants who clustered with the US-majority Model 2(n = 8) diverged from their Ngöbe Model 1 counterparts in theirlower attributions of agency to plants (on every capacity exceptmother tree helping), as well as their lower endorsement ofmachine agency capacities. However, they remained distinct fromUS members of Model 2 by being more likely to endorse a plant’scapacity to feel another tree’s pain. Ngöbe and US Model 2 mem-bers did not reliably differ on any other measures of plant agency,indicating a fair amount of convergence between this subgroup ofNgöbe participants and the US-majority model.

The two US participants who clustered with the Ngöbe-majorityModel 1 differed from their US Model 2 counterparts in theirhigher endorsement of plant agency (for kin recognition, and allfive intentional agency capacities: see, feel, pain, think, want). Theyremained distinct from Ngöbe members of Model 1, however, bybeing less likely to endorse plants’ capacity to experience pain orfeel another tree’s pain.

These findings lend additional evidence for multiple models ofagency among Ngöbe respondents. The focal model among Ngöberespondents presents a strong recognition of plant agency, but thisis accompanied by pronounced variation towards an animal-oriented model where plant agency is diminished. For their part,the rare US individuals who departed from their group’s dominantanimal-oriented model of agency moved towards an increasedrecognition of plant intentional (more so than social) agency.

Individual differences in competence. One unique strength of CCMis that it goes beyond sample means to assess each individual’slevel of competence on their consensus model, or how well theyfit the consensus model to which they were assigned. We next ana-lyzed individual differences associated with these competencescores. Analyses were run separately for Model 1 and Model 2scores within each cultural group (Ngöbe or US).

For Ngöbe participants, we anticipated that greater competenceon the plant-oriented Model 1 would correlate with more tradi-tional community-based lifestyles, whereas competence on theUS-majority Model 2 would show the inverse. However, no reliablerelationships emerged between competence and lifestyle charac-teristics for Ngöbe clustered with either Model 2 (n = 8) or Model1 (n = 9), likely due to modest sample sizes.

For US participants clustered with the US-majority Model 2, theonly individual characteristic that reliably correlated with compe-tence was, unexpectedly, gender, such that males were more com-petent than females (r = 0.54, p < 0.01). It is unclear why this maybe the case.

Further investigation showed that competence among US par-ticipants exhibited interesting conceptual patterns. US members’


competence on Model 2 was associated with more predictions forkin altruism but fewer predictions for communication, whereasneither inference task predicted US membership in Model 1 versusModel 2—instead, membership on Model 1 was associated withgreater plant agency attributions. This suggests that plant agencyconcepts break down into separate components that may track dif-ferent individual characteristics among US participants. We pur-sued this by exploring individual characteristics associated withplant agency measures among the entire US sample (n = 30). Inter-estingly, religious belief correlated with greater willingness toascribe agency capacities to plants (specifically, thinking and want-ing), F(1,25) = 4.65, p < 0.05, g2 = 0.16, whereas growing up in anonurban environment correlated with greater predictions forplant social interaction (specifically, communication), F(1,25)= 4.16, p = 0.05,g2 = 0.14. These separate patterns for US religiosityand childhood residence suggest that knowledge of plant behaviormay track separately from interpretive frameworks for under-standing that behavior.

10.4. Cultural modeling: discussion

Cultural consensus modeling allowed us to identify whethermultiple conceptual models of agency are present among ourrespondents using a bottom-up modeling approach that treats cul-ture as a latent variable. Two distinct consensuses were detected,confirming that there are two qualitatively distinct conceptualmodels of agency. One model endorses an animal-oriented viewof agency where plants possess a very constrained set of capacitiesfor signaling and sensation (endorsed by the majority of US respon-dents and nearly half of Ngöbe respondents). But an alternativemodel also stands, one that takes a decidedly different view whereplants possess a robust suite of agency capacities on par with ani-mals (endorsed by a majority of Ngöbe informants and two USrespondents). These results provide evidence against the claim thatall individuals share a basically animal-oriented model of agency,while also highlighting that multiple conceptual models are avail-able within a given cultural context.

Investigation of the individual characteristics aligned withthese models further illuminated the sources of variation underly-ing diverse conceptual stances. The results suggest that amongNgöbe respondents, the plant-oriented model of agency is sup-ported by engagement with Indigenous cultural epistemologies.Among US individuals, religious belief and nonurban experiencewere correlated with distinct aspects of plant agency concepts.Religious belief was correlated with greater capacity attributionsto plants, whereas nonurban experience correlated with greaterpredictions for plant social interactions. This suggests that episte-mological orientations (e.g., religion) may influence interpretativeframeworks for plants while familiarity with the biological worldmay inform inferences about actual plant behavior.

11. General discussion

11.1. Summary

Across two experiments using mixed methods, we found thatNgöbe were more likely than US individuals to ascribe a widerange of agency capacities to plants and to predict that plants willengage in complex social behaviors. Results from bottom-up mod-eling revealed that these cultural differences derive from two qual-itatively distinct conceptual models of agency rather thanvariations on a single model. Furthermore, the present findingsare unlikely to reflect cross-linguistic differences, as task languageaffected neither capacity attributions nor consensus models ofagency.

Two key conclusions can be drawn from these results. First, theNgöbe consensus model points to a framework that interpretsplant and animal actions on closely corresponding terms. Thischallenges the domain-specificity claim that plants are universallyconceptualized as insentient objects that belong to a differentdomain than animate agents. Second, the Ngöbe view correlateswith insightful inferences about plant social behaviors that haveonly recently been observed by Western plant scientists(Mancuso & Viola, 2015). This suggests that the Ngöbe perspectiveon plants as active agents may afford insights into complex plantbehaviors that are less accessible on the US view of plants as insen-tient, passive biological kinds.

Clearly, many Ngöbe informants conceptualize plants as agentsin a way that US individuals do not. The critical question is how tounderstand the conceptual commitments at stake in theseobserved cultural differences. In what follows, we consider threepossible accounts of the present findings in terms of (1) differentdegrees of ecological expertise, (2) extensions of universal folkpsy-chology concepts, or (3) culturally variable concepts of agency.

11.2. Ecological expertise

One might argue that Ngöbe inferences for plant social behav-iors can be explained by their greater ecological knowledge ascompared to US college students, rather than resulting from dis-tinct conceptual frameworks per se. Certainly Ngöbe can beexpected to have considerable knowledge of plants through theirdaily interactions with the surrounding forest ecology. However,Ngöbe informants’ predictions for plant communication and kinaltruism cannot be explained solely by expertise because thesespecific plant social interactions represent recent discoveries inWestern botanical science. If sheer quantity of experience withplants led Ngöbe to observe subtle signaling and kin resource shar-ing behaviors, then this begs the question of why those insightslong eluded Western plant scientists. We suggest that the Ngöbeconceptual framework provides a lens for expecting complexbehavior which guides their inferences even in novel situationssuch as these. Converging evidence that cultural frameworks guideinferences for novel cases of nonhuman behavior comes fromanother study where we found that Ngöbe are more likely thanUS individuals to accurately infer cooperation between two animalpredators that are unfamiliar to them (ojalehto, Medin, Horton,Garcia, & Kays, 2015).

It is important to ask how ecological expertise and concepts ofnonhuman agency may interact, but the answer is unlikely to be assimple as expertise versus concepts. We see interpretative frame-works for plant agency (whether encoded in folktheories or scien-tific theories) as important conceptual devices for organizingknowledge and integrating new information (Hall, 2011; Narby,2006), as well as guiding inference and explanation (Keller, 1983;Medin & Bang, 2014). From this perspective, conceptual frame-works necessarily involve feedback with many forms of culturalexperience including ecological knowledge. It is informative toconsider howWestern scientists have brought their own intuitionsto the study of plant behavior. For instance, initial findings forplant kin recognition were met with skepticism among plant scien-tists and some even declared (dubiously) that it would be ‘‘extraor-dinary” if the early claims could actually be substantiated(Klemens, 2008, p. 67). Those claims have since been substantiatedto the satisfaction of many in the field (Bhatt et al., 2011;Biedrzycki & Bais, 2010). Yet the initial skepticism serves to high-light how even experts’ inferences concerning plant behaviormay reflect prior assumptions—ecological expertise is not indepen-dent of cultural intuitions. Our results serve as a critical first proofthat those cultural intuitions do differ.


Our findings align with broader interdisciplinary research pro-grams demonstrating that Indigenous perspectives contributenovel insights into biological and ecological structure (Anderson,2012; FORMABIAP, 2008; Narby, 2006; Pierotti, 2011; Pitmanet al., 2011; Waldstein & Adams, 2006). Such observations andour own results raise important questions concerning the interac-tions between ecological expertise and cultural epistemologies asthey relate to concepts of nonhuman agency and environmentalcognition more broadly (e.g., ojalehto, Medin, Horton, Garcia, &Kays, 2015; ojalehto, Waxman, & Medin, 2013). With cultural dif-ferences now established by the present experiments, further workmay begin exploring how varying levels of expertise within eachcultural context may (or may not) influence conceptions of naturalagents.

11.3. Folkpsychology

Our results could be interpreted as showing that Ngöbe overex-tend a folkpsychological concept of mind to plants (e.g., Atran &Norenzayan, 2004; Barrett, 2000; Boyer, 1996, 2003). This overex-tension account would rest on the assumption that Westernfolkpsychology is a universal conceptual framework whose properdomain is restricted to animate agents, defined as humans and ani-mals (e.g., Sperber & Hirschfeld, 2004; Wellman & Gelman, 1992).On this perspective, Ngöbe are incorrectly inferring human-likemental states on the part of plants. If the current results are inter-preted in this way two questions follow. First, if the Ngöbe concep-tion of plant agency affords useful insights into plant behavior thenis it appropriate to characterize it summarily as a mistaken infer-ence? Related to this, we must also consider why the US folkpsy-chological stance leads to the misguided rejection of actual plantcapacities for social behavior.

Considering these questions, we propose that US attitudesabout plants may be interpreted as an under-extension of agencyconcepts that are constrained by cultural beliefs about brains andnervous systems (for related points, see Marder, 2012, 2013). Theidea that anthropocentric folkpsychology represents a selectivenarrowing of the agency domain would be consistent with severallines of evidence. First, based on infancy studies, it has been sug-gested that the extension of the concept ‘‘agent” is defined by prin-ciples of interaction (e.g., goal-directedness) and has no properontological domain (Gergely & Csibra, 2003; see also Lowder &Gordon, 2015). This has led some to propose that the core repre-sentational system for agency actually has no proper intuitiveontological domain: ‘‘Our tendency to apply psychological expla-nations to non-human phenomena...may not be due to an overex-tension of an (originally more restricted) psychologicalinterpretational framework,” but may instead reflect the properapplication of a teleological stance (Csibra, Gergely, Bıró, Koos, &Brockbank, 1999, p. 265). (We agree but hasten to add that a psy-chological framework need not be the only possible one for under-standing teleological nonhuman action.) Given that plants exhibitthe critical cues for agency perception (e.g., goal-directedness, con-tingent interaction), they would also fall within the proper applica-tion of this teleological stance.

Furthermore, the evidence suggests that culture plays a role inacquiring an anthropocentric folkpsychology constrained tohumans and animals. A large body of work in developmental psy-chology indicates that infants possess a core representational sys-tem for agency online by the end of the first year of life (Bíró &Leslie, 2007; Csibra & Gergely, 2013; Johnson, 2003; Johnson,Shimizu, & Ok, 2007; Woodward, 2013). However, this is a skeletalrepresentation system and it is only later on that children developa rich conceptual framework for agency that includes causal prin-ciples tied to minds and mental states, where behavior is under-stood on folkpsychological terms (Johnson, 2000; Wellman,

2010). For US children, learning that psychological propertiesshould not be extended to all actors who exhibit goal-directedbehavior (such as plants or other inanimate nonhuman kinds)but rather selectively applied to humans and other animalsrequires extended learning into the grade school years (Opfer,2002). This indicates that the exclusion of plants from the domainof agency may be a culturally acquired stance.

In short, the developmental evidence is consistent with the pos-sibility that anthropocentric folkpsychology represents a culturallyacquired lens on agency that narrows the domain to animates onthe basis of a human prototype. The corresponding folkpsycholog-ical notions of minds and mental states may reflect a culturalmodel among Western researchers and participants (for similarconclusions in related domains, see Medin, ojalehto, Waxman, &Bang, 2015). This raises the possibility that the infant’s representa-tional system for agency—while widely seen as a precursor tofolkpsychology among Western samples—may serve as a precursorto alternative folktheories of agency in other cultures. This is espe-cially plausible when we consider that capacities which cue agencydetection (goal-directed motion, contingent interaction) are notexclusive to entities with human-like brains but are also carriedout by plants and other organisms (Chamovitz, 2012; Marder,2012; Uexküll, 2010). Yet by focusing on the specifically psycho-logical aspect of agency as it is assumed on Western folktheories,researchers have rarely considered the possibility that culturalfolktheories of agency may be organized around other, non-psychological concepts (for a notable exception in the domain offolksociology, see Hirschfeld, 2006, 2013). Considering this possi-bility allows alternative accounts to be offered.

11.4. Folkcommunication

It is possible that the current findings point to a distinctive con-ceptual framework for agency that is available to Ngöbe infor-mants. Specifically, we have argued that Ngöbe endorsements ofplant agency can be understood in terms of folkcommunication,where the capacity for relational interaction is the grounds forinferring agency. This framework would properly apply to diverseactors including (at least) plants and animals. This account locatescultural variation at the level of conceptual organization of folkthe-ories rather than in variable extensions of universal folkpsycholog-ical concepts.

Folkcommunication is at least as compatible with the data as anexplanation based on overextensions of folkpsychology. Further-more, folkcommunication carries the advantage that it makesfewer assumptions about mental-state concepts presumably heldby Ngöbe, and avoids the explanatory paradox of why Ngöbewould consistently make ‘‘category errors” concerning plantsdespite having considerable ecological experience. On folkcommu-nication, plant agency attributions are seen as an empiricallygrounded recognition of communicative exchanges among diversenatural kinds.

We propose that folkpsychology and folkcommunication, whilelikely deriving from a common early-developing cognitive systemfor detecting agency (Csibra et al., 1999; Johnson, 2003), follow dis-tinctive meaning-making principles that ultimately lead to differ-ent domains. This need not imply that the two folktheories aremutually exclusive, such that US individuals have no access to folk-communication, for instance. Indeed, our cultural consensus mod-eling results show that both models are present in the Ngöbecontext and possibly in the US context as well. We speculate thateither conceptual framework may be cultivated to a differingdegree in various contexts. Even within a single individual, thesefolktheories could represent interpretive stances that focus on dis-tinct properties of agency. On folkcommunication, the focus is onrelational interactions as the primary quality of agency, and mental


states may play a secondary role if any in agency judgments. Bycontrast, the folkpsychological lens trains attention to mentalstates as the primary property of agency and if communication fig-ures in this model, it may be seen as a derivative property of themental (Dennett & Haugeland, 1987).

Our account is roundly consistent with evidence from relateddomains to suggest that Western anthropocentrism leads to pat-terns of reasoning that differ from a more ecological orientationamong Indigenous communities (Atran, Medin, & Ross, 2005;Bang, Medin, & Atran, 2007; ojalehto et al., 2013; Waxman &Medin, 2007), and resonates with other research proposing thatbroad framework principles in the folkpsychological domain mayrecruit different bodies of socially learned knowledge across cul-tures (Meltzoff & Gopnik, 2013, see also Gelman, 2009). It is alsowell supported by a growing body of anthropological evidenceon Indigenous folktheories of communication in ecosystems (seeespecially Kohn, 2013, 2015).

The current experiments are an initial foray and many ques-tions remain. To date, folkpsychology has been virtually the onlyframework in town for thinking about agency, so it requires futurework to clarify the inferential principles and explanatory con-structs at work in folkcommunication. One line of investigationcould target cultural differences in attributions of predicates andinferences that might distinguish between folkpsychology (e.g.,internal individual states) and folkcommunication (e.g., interactiverelational states). Cultural diversity may arise from differences inthe relative salience given to these conceptual nodes and theirextension to different classes of agents (e.g., plants, abiotic kinds,social groups, artificial intelligence).

12. Conclusion

Our research project sought to explore variability in concepts ofnonhuman agency, specifically plant agency, and to generate newforms of evidence that can speak to the conceptual commitmentsat stake in animism. The experimental results show substantialcultural variation in concepts of plant agency and demonstrate thatthe Ngöbe perspective affords sophisticated insights into plantbehavior. Our converging methods extend prior work by observingsystematic differences in nonhuman agency attributions acrosscultures, establishing that these cultural differences extend toinferential reasoning, and showing that these cultural patternsarise from two distinct conceptual models of agency rather thanvariations on a single universal model. The differences observedbetween US and Ngöbe informants may arise from different exten-sions of a common folkpsychology, or from a distinctive theory offolkcommunication among Ngöbe informants. The present resultsare suggestive of the latter possibility but it remains for futureresearch to disentangle these accounts.

At the very least, these findings raise the possibility that domain-specific conceptual knowledge may be organized in ways that areculturally variable. The anthropocentric stance on folkpsychologyas thekeyorganizing framework for agencymaybe a featureofWes-tern cultural epistemologies rather than a universal cognitive prin-ciple. As an alternative, folkcommunication offers an ecologicallyoriented lens for attending to complex forms of interaction amongnonhuman agents. The Ngöbe perspective serves as an invitationto revisit the widespread assumption among Western researchersthat mentalistic folkpsychology is the universal intuitive prototypefor agency—and that plants belong outside this domain. Pursuingthe invitation to reconsider anthropocentric views of agency couldadvance cognitive theorywhile expandingWestern understandingsof nonhuman actors. The question of how humans conceptualizenonhuman beings in their ecological communities has never beenmore relevant than it is today.

Acknowledgements

We thank the Ngöbe community for their support of and partic-ipation in this research. We are grateful to Eugene Anderson,Daniel Casasanto, Rumen Iliev, Joshua Knobe, Jonas Nagel, JeremyNarby, Ara Norenzayan, Jeremy Ojalehto, Lance Rips, Sonya Sach-deva, Rebecca Seligman, Richard Shweder, Sandra Waxman, andtwo anonymous reviewers for helpful discussion of these ideas.Thanks to John Opfer for generously sharing the materials usedin Experiment 1, and to Royce Anders for sharing his expertiseon cultural consensus modeling. For their contributions to data col-lection and analysis we thank Linda Flores and Yereida Gallardo.Portions of this research were presented at the Cognitive ScienceSociety Meeting in Pasadenda, CA in July, 2015; the Society forAnthropological Sciences Meeting in Pittsburgh, PA in March,2015; and the University of Chicago Cognitive Brownbag in Febru-ary 2015. The research and writing of this article was supported bythe National Science Foundation under Grant Numbers NSF DRMS1427035 to bethany ojalehto and Doug Medin, SES0962185 andDRL1114530 to Douglas Medin, and an NSF Graduate Research Fel-lowship to bethany ojalehto.

Appendix A. Experiment 1

A.1. Experiment 1A: supplementary methods and materials

Culture-specific stimuli for plants and machines were providedin Panama, as follows:

Plant item (Benefit Present condition): Here is a plant thatgrows on the ground. This kind of plant depends on sunlightto grow and live. The seedling is starting to grow in the shade,here. The sunlight is over there. When this plant grows, will itkeep growing straight forward in the shade, or will it grow overto where the sunlight is?Aquí hay una planta que crece en el suelo. Esta planta necesita luzdel sol para poder crecer y vivir. La semilla esta empezando a creceraquí, en la sombra. La luz del sol esta justo ahí. ¿Cuándo estaplanta crezca, se mantendrá creciendo recto debajo del sombra, ocrecerá hacia donde está la luz?

For machines, we used line drawings of three familiar machines(motorboat, radio, and chainsaw) to replace the original novelmachines (N = 10). Otherwise, all aspects of the machine stimuliwere conserved (i.e., locomotion, goal objects). (Culture-specificmachine items are available from the first author upon request.)

A.2. Experiment 1B: supplementary methods and materials

Question probes for capacity attributions as translated into eachstudy language (English, Spanish, and Ngöbere) are providedbelow, using as an example item the plant growing toward sun-light. Each probe was phrased in relation to the goal object (e.g.,can a plant see, want, feel, or think about the sun?), with theexception of pain, which specified a form of harm (e.g., cutting aplant). As such, each capacity attribution represented an interpre-tation of that entity’s specific goal-directed action presented in thetask. This method follows that used in Opfer & Gelman’s originalstudy, with the exception of wanting (which was phrased withthe object ‘‘anything” rather than the goal object).

Can the plant see the sunlight?¿La planta puede ver la luz del sol?Ngwina trö törö köteri gwane, dba töin iye ya? (Si hay luz del soladentro, ella puede ver la luz?)

Can the plant feel the sunlight when it touches it?

Table A1Mean attributions of agency capacities by culture and kind for Experiment 1B.

Psychological capacity/Culture Animals Plants Machines Artifacts Capacity Total

Mean SD Mean SD Mean SD Mean SD Mean SD

FeelNgöbe 0.97 0.10 0.84 0.36 0.33 0.37 0.23 0.31 0.59 0.29US 0.98 0.08 0.69 0.32 0.16 0.31 0.02 0.08 0.46 0.20Total 0.98 0.09 0.75 0.34 0.23 0.35 0.11 0.23 0.52 0.25

WantNgöbe 1 0.00 0.78 0.37 0.35 0.41 0.13 0.28 0.56 0.26US 0.95 0.12 0.37 0.47 0.13 0.30 0.01 0.06 0.36 0.24Total 0.97 0.10 0.54 0.48 0.22 0.36 0.06 0.19 0.45 0.28

PainNgöbe 0.97 0.10 0.71 0.44 0.16 0.35 0.10 0.25 0.49 0.28US 1 0.00 0.08 0.22 0.00 0.00 0.01 0.06 0.27 0.07Total 0.99 0.06 0.34 0.45 0.07 0.23 0.05 0.17 0.36 0.23

ThinkNgöbe 0.83 0.39 0.46 0.50 0.09 0.29 0.07 0.24 0.36 0.36US 0.94 0.24 0.16 0.36 0.08 0.24 0.00 0.00 0.30 0.21Total 0.89 0.31 0.29 0.45 0.08 0.26 0.03 0.16 0.32 0.29

SeeNgöbe 1 0.00 0.26 0.41 0.04 0.21 0.04 0.21 0.34 0.21US 0.91 0.21 0.01 0.06 0.05 0.19 0.00 0.00 0.24 0.11Total 0.95 0.17 0.11 0.29 0.05 0.20 0.02 0.13 0.28 0.20Kind Total 0.95 0.13 0.43 0.37 0.14 0.27 0.06 0.16 0.39 0.23

Overall psychological attributions (composite scores averaged across 5 capacities) Total

Ngöbe 0.95 0.08 0.61 0.33 0.19 0.25 0.12 0.22 0.47 0.18US 0.96 0.07 0.26 0.20 0.08 0.18 0.01 0.02 0.33 0.08Total 0.96 0.08 0.40 0.31 0.13 0.22 0.05 0.15 0.39 0.15

Table A2Table of effects for Exp. 1B: Culture, kind, and capacity attributions.

Effect F test Post-hoc

Culture F(1,54) = 15.26*** Ngöbe (M = 0.47) > US (M = 0.33)g2 = 0.22

Kind (3,162) = 388.53*** Animals (M = 0.95) > Plants (M = 0.43) > Machines (M = 0.14) > Artifacts (M = 0.06)g2 = 0.88

Capacity attribution F(4,216) = 40.78*** (Feel, Want > Pain, Think, See) (Pain > See)g2 = 0.43 Feel (M = 0.52), Want (M = 0.45), Pain (M = 0.36), Think (M = 0.32), See (M = 0.28)

Culture � Kind F(3,162) = 13.09*** Plants: Ngöbe (M = 0.61) > US (M = 0.26)g2 = 0.20 Artifacts: Ngöbe (M = 0.12) > US (M = 0.02)

Culture � Capacity F(4,216) = 4.36* Feel: Ngöbe (M = 0.59) > US (M = 0.46)g2 = 0.08 Want: Ngöbe (M = 0.56) > US (M = 0.36)

Pain: Ngöbe (M = 0.49) > US (M = 0.27)See: Ngöbe (M = 0.34) > US (M = 0.24)

Kind � Capacity F(12,648) = 16.04*** Animals > All other kinds: for all 5 capacitiesg2 = 0.23 Plants > Machines, Artifacts: for 4 capacities (not See)

Machines > Artifacts: for WantCulture � Kind � Capacity F(12,648) = 4.59*** (see separate tests for each kind below)

g2 = 0.08

Animals:Culture � Capacity F(4,216) = 2.96* See: Ngöbe (M = 1) > US (M = 0.91)

g2 = 0.05

Plants:Culture � Capacity F(4216) = 5.57** Feel: Ngöbe (M = 0.84) > US (M = 0.69)

g2 = 0.09 Want: Ngöbe (M = 0.78) > US (M = 0.37)Pain: Ngöbe (M = 0.71) > US (M = 0.08)See: Ngöbe (M = 0.26) > US (M = 0.01)

Machines:Culture � Capacity F(4,216) = 3.40* Want: Ngöbe (M = 0.35) > US (M = 0.13)

g2 = 0.06 Pain: Ngöbe (M = 0.16) > US (M = 0.00)

Artifacts:Culture � Capacity F(4,216) = 4.97** Feel: Ngöbe (M = 0.23) > US (M = 0.02)

g2 = 0.08 Want: Ngöbe (M = 0.13) > US (M = 0.01)

* P-value > 0.05.** p-value > 0.01.*** p-value > 0.001.



¿La planta puede sentir el sol cuando la toca?Ngwina trö törö köteri gwane, dba doin iye ya? (Si hay luz del soladentro, ella puede sentir la luz?)

If it’s cut, can the plant feel pain?¿Si se corta, esta planta puede sentir dolor?Mö we dö ködrete jödr _onbidi gwane, dba nme tare iye ya? (Siusted la puyas con algo, ella puede sentir dolor?)

Can the plant think?La planta puede pensar?Dba töbike? (Ella puede pensar?)

Can the plant want the sun?¿La planta puede querer el sol?Ngwina trö törö köteri gwane, dba tönme iye ya? (Si hay luz delsol adentro, ella puede querer la luz?)

A.3. Experiment 1B: supplementary results

A.3.1. Missing item responseDue to experimenter error, six US participants were presented

with the ‘‘want” question for only 2 of 3 plants. Their average‘‘want” attributions to plants were collapsed over 2 rather than 3items (this was warranted as responses were uniform for bothplant items for all 6 participants).

A.3.2. ResultsIn addition to the cultural differences reported in the main text,

we found the expected main effects and interactions (as evident inFigs. 1and 2 presented in the main text); descriptive statistics areprovided in Table A1 and tests are reported in Table A2. Maineffects: There were main effects of kind (animals > plants > machi-nes > artifacts, all ps < 0.01) and of capacity (feeling and wantingwere the most frequent attributions, ps < 0.01, and pain more thanseeing, p < 0.01). Condition: As expected, BP/BA condition had noeffect on psychological attributions for either group.

A.3.3. InteractionsIn addition to cultural differences in capacity attribution to

plants (see main text), cultural differences of lesser magnitude alsoemerged for the three other kinds. Animals: Ngöbe were morelikely than US participants to attribute seeing to animals. Inspec-tion of US explanations indicated that situational features of taskscenarios prompted infrequent denials that an animal could seethe goal object (one was in the animal’s mouth; another was insidea log). Machines: Ngöbe were more likely than US participants toattribute pain and wanting to machines. Artifacts. Ngöbe were alsomore likely than US participants to attribute feeling and wanting toartifacts. These cultural differences were minor in the absolutesense, and attributions to machines and artifacts were low evenamong Ngöbe.

A.3.4. Ngöbe culture-specific plant stimulusAmong Ngöbe participants, capacity attributions to plants

unexpectedly differed across the two stimulus versions, such thatparticipants were more likely to attribute agency capacities tothe culture-specific item depicting an outdoor ground vine(M = 0.86, SE = 0.09) than the original study item from Opfer andGelman (2001) depicting an indoor potted plant (M = 0.45,SE = 0.08), F(1,21) = 11.44, p < 0.01, g2 = 0.35). As we found forExp. 1A goal-directed action predictions, however, this effect heldfor all three plant stimuli (not just the item that differed across ver-sions), so other factors (e.g., participant schooling and churchmembership) contribute to the observed difference.

Appendix B. Language

B.1. Language protocol

Primary probes were always presented in the assigned inter-view language but participants were free to respond in their pre-ferred language; nearly all responded in the assigned interviewlanguage (including the US text-based responses, and Panama ver-bal responses).

B.2. Supplementary language results

Among US participants, there were no reliable effects ofSpanish versus English for any of the experiments. Among Ngöbeparticipants, no language effects were observed for Spanish ver-sus Ngöbere on the capacity attribution tasks (Experiments 1Band 2B), where any cross-linguistic differences could have beenrelevant as an alternative account of observed culturaldifferences.

Unexpectedly, Spanish-Ngöbere language differences werefound among Ngöbe participants for the goal-directed action pre-diction task (Experiment 1A). Participants predicted more goal-directed actions for plants (across both BA and BP conditions)and animals in Ngöbere than Spanish (Plants: F(1,19) = 6.46,p < 0.05, g2 = 0.25; Ngöbere M = 0.72, SD = 0.28; Spanish M = 0.44,SD = 0.40) (Animals: (1,19) = 4.67, p < 0.05, g2 = 0.20, NgöbereM = 0.97, SD = 0.10; Spanish M = 0.73, SD = 0.42). Predictions formachines and artifacts did not differ. This is the one study wherewe did not predict (or find) cultural differences between US andNgöbe samples, so the observed language effects are incidental toour primary cultural hypotheses. In any case, they do not alterthe overall findings reported for Experiment 1A. We do not thinkthis indicates that Spanish instructions were unclear to partici-pants, because action predictions did not differ for machines orartifacts by language.

The observed effects of language may point to other factors.Both language and stimulus version were correlated with demo-graphic characteristics in our sample: participants in the Ngöberecondition tended to be older and to have fewer years of formalschooling, while participants in the familiar plant stimulus versionhad less schooling (correlated with stimulus version r = �0.62,p < 0.01) and were more likely to be members of the IndigenousNgöbe Church (r = 0.59, p < 0.01). The multiply correlated factorsin our sample make it difficult to isolate a single factor in explain-ing the observed differences for predictions of plant goal-directedactions.

Appendix C. Experiment 2

C.1. Participants, pilots and protocol development

Development of the action prediction scenarios and capacityquestions involved piloting with several participants. Addition-ally, some participants did not complete the interview, eitherleaving blank survey responses (in the US) or finishing early toavoid exhaustion in longer interviews (in Panama). Altogetherincluding pilots, 22 Ngöbe and 32 US individuals participated inExperiments 2A and 2B. Given that many items that pilot partic-ipants saw were nearly identical to the final items, we also con-ducted all analyses reported here on the full participant sampleincluding pilots. All findings hold across both sample sets (seeTable C1).

Table C1Coding categories for explanations of plant social action predictions.

Code Domain/Example explanations

Other-directed goalCommunication

‘‘They were communicating underground and telling the other tree,they were warning it that it has to save some water.” (Ngöbe,Drought)‘‘The plants worked synergistically, together, so that they could all beprotected.” (US, Insects)

Kin altruism‘‘Because the mom is giving her food to eat. . .as if the mom is givingher some help to grow.” (Ngöbe, Mother tree)‘‘The mother tree will help its baby via roots.” (US, Mother tree)

Individual goalCommunication

‘‘Maybe they feel the chemical defense on the other side and that’swhy they think that it is better that they prepare too.” (Ngöbe, Insects)‘‘They sense the chemicals and it’s a natural response to survive. Theroots were very close.” (US, Insects)

Kin altruism‘‘It will be next to its mom and it will grow better because if you put itapart, then it will starve.” (Ngöbe, Mother tree)‘‘It can take nutrients from its mother to help it grow instead of havingto completely fend for itself.” (US, Mother tree)

MechanisticCommunication

‘‘Because it will be spreading its roots, its leaves, its flowers. It isstrong. (Ngöbe, Insects)‘‘Its roots will likely have been changed due to the other plant. It’s not‘preparing itself’, it’s responding to a simple stimulus from anotherplant to change.” (US, Drought)

Kin altruism[No]a ‘‘The territory where the roots are, the mother tree, is alreadytoo dry. The nutrients that are there are not strong enough for a newseed to grow.” (Ngöbe, Mother tree)‘‘The roots help stabilize the young tree.” (US, Mother tree)

a All explanations are for positive action predictions, except this case where nonewas available because no Ngöbe respondent provided a mechanistic explanation forkin altruism.


C.2. Experiment 2A: supplementary methods and materials

Communication scenarios

(i) Insect communication scenario

Here are 3 plants of the same species. There is an above-ground wall separating the plant on the left from the two plantson the right. The leaves of the individual plants can’t touch eachother, but the roots are connected. Some insects have arrivedand are irritating the plant on the left. As a result, the planthas released chemicals to defend itself from the insects. Thechemical is released through both the leaves and the roots.What will happen to the two plants on the other side of the wall,after the left plant releases defensive chemicals against theinsects? (Assume that the insects have not arrived on the right

side of the wall yet.) (1) The other two plants will also releasetheir chemicals to defend themselves. (2) The other two plantswill remain in the same state as before, without releasingchemicals.

Aquí hay 3 plantas de la misma especie. Hay una pared encima delsuelo separando la planta a la izquierda de las otras dos plantas a laderecha. Las hojas de las plantas individuales no pueden tocarse lasunas con las otras, pero las raíces si están conectadas. Han llegadounos insectos a molestar la planta a la izquierda. A consecuencia, laplanta ha echado unos químicos para defenderse de los insectos. Losquímicos son descargados por medio de las hojas y las raíces. ¿Quéva a pasar con las dos plantas al otro lado de la pared después quela planta a la izquierda echo químicos defensivos contra los insectos?(Asume que los insectos todavía no han llegado al lado derecho de lapared.) (1) Las otras 2 plantas van a echar el químico para defendersede los insectos también. (2) Las otras 2 plantas se van a quedar igualcomo siempre, sin echar químicos.

(ii) Drought communication scenario

Here are three plants of the same species. These plants havelong roots that extend to great depths because they live in an aridarea where water is located deep underground. There is one plantgrowing by itself on the left side of a huge rock cliff. It is con-nected to the middle plant via roots that travel beneath the rock,but it is not connected to the third plant on the far right. The cli-mate on the left side of the cliff is a bit different from the climateon the right side. Currently, there is a severe drought on the leftside. In order to defend itself from the drought, the left plantundergoes a change in its body that allows it to conserve itswater use. The drought has not yet reached the other side ofthe cliff, but it will arrive there in three weeks with even moreintensity. What will happen to the other two plants when thedrought arrives on the right side of the cliff? (1) They are bothgoing to die because neither of them were prepared beforehand.(2) Only the plant that was connected will survive because itwas prepared beforehand.

Aquí hay tres plantas de la misma especie. Estas plantas tienenraíces muy largas que alcanzan una profundidad porque viven en unárea seca donde el agua está profundamente debajo del suelo. Hayuna planta que está creciendo sola al lado izquierdo de una rocagrande. Esta planta está conectada a la planta en medio por mediode las raíces que recorren debajo de la roca, pero no está conectadaa la tercera planta que está a la derecha. El clima a la izquierda dela roca es un poco diferente del clima al lado derecho de la roca. Ahor-ita, hay una sequía grande al lado izquierdo. Para defenderse de lasequía, la planta a la izquierda hace un cambio en su cuerpo que lapermite conservar su uso de agua. La sequía todavía no ha alcanzadoel lado derecho de la roca, pero va a llegar en tres semanas con másintensidad. ¿Qué va a pasar con las otras dos planta cuando lleguela sequía al lado derecho de la roca? Que pasara cuando la sequía lle-gue donde ellas? (1) Ellas dos van a morir porque ninguna de ellasestaban preparadas. (2) Solo la planta que estaba conectada va asobrevivir porque ella estaba preparada antes.

Kin Altruism Scenarios


(iii) Sibling kin altruism scenario

The two plants in this pot are of the same species, but are unre-lated (not biological kin). If one of the plants is stronger than theother and can more quickly extend its roots, it will capture allthe nutrients until the other plant dies. This frequently occursamong unrelated plants. Now we have two plants that are related:they are biological kin. One of the plants (on the left) is strongerthan the other and can extend its roots faster than its sibling (onthe right). What will happen in this case with the two plants thatare biological kin? (1) The stronger plant will extend its roots inall directions and take all the nutrients until its sibling dies. (2)The stronger plant will share the nutrients with its sibling and bothwill survive.

Estas dos plantas en la maceta son de la misma especie, pero noestán relacionadas (no son descendentes de la misma planta). Si unade las plantas es más fuerte que la otra y puede crecer sus raíces entodas partes más rápidamente, puede capturar todos los nutrienteshasta que la otra planta se muera. Esto pasa con frecuencia entre plan-tas que no son relacionadas. Ahora tenemos dos plantas que son rela-cionadas: son descendentes de la misma planta. Una de las planta (ala izquierda) es más fuerte que la otra y puede extender sus raícesmás rápido que su hermano (a la derecha). ¿Que pasara en este casocon las dos plantas que son descendentes de la misma planta? (1) Laplanta más fuerte va a crecer sus raíces en todas direcciones y capturartodos los nutrientes hasta que su hermano muera. (2) La planta másfuerte va a compartir los nutrientes con su hermano y los dos van asobrevivir.

(iv) Mother tree kin altruism scenario

Here is a large tree that is growing in the mountains. Afterreproducing, it releases one of its seeds. Where will the seed growbetter? (Suppose that there is the same amount of sunlight andnutrients in both locations.) (1) Near its mother’s roots. (2) Fartheraway where there are no roots from its mother.

Aquí hay un árbol grande que está creciendo en la montaña.Después de haber reproducido, el árbol descargo una de sus semillas.Dónde va crecer mejor la semilla? (Supone que hay la misma cantidadde luz del sol y nutrientes en las dos posiciones.) (1) Cerca de las raícesde la mama. (2) Más lejos donde no hay raíces de la mama.

C.3. Supplementary results on plant interaction predictions

C.3.1. Plant communicationSeparate analyses of each prediction scenario showed that this

cultural difference held for both communication in response toinsects (Ngöbe M = 0.94, SD = 0.24; US M = 0.70, SD = 0.47), t(44.80) = 2.33, p < 0.05, d = 0.70, and communication in responseto drought (Ngöbe M = 0.94, SD = 0.25; US M = 0.70, SD = 0.47), t(43.98) = 2.25, p < 0.05, d = 0.68. Kin altruism: Separate analyses ofeach prediction item showed thatNgöbewere farmore likely to pre-dict sibling resource sharing (M = 0.82, SD = 0.39) than US partici-pants (M = 0.30, SD = 0.47), t(45) = 3.91, p < 0.001, d = 1.16, butthere was no difference in predictions for mother tree, t(43)= �0.03, p = 0.98 (NgöbeM = 0.35, SD = 0.49;USM = 0.36, SD = 0.49).

C.4. Experiment 2A: supplementary details on coding analysis

C.4.1. Coding reliabilityTwo independent coders judged a subset of explanations

(representing over 30% of the corpus of explanations) for interpre-tations of plant interactions. Inter-rater reliability on the assign-ment of explanatory codes was high for both communicationexplanations (Kappa = 0.81, p < 0.001; percent agreement = 88%)

and kin altruism explanations (Kappa = 0.76, p < 0.001; percentagreement = 88%).

C.4.2. Detailed coding results

Kin altruism: Explanation data was available from US partici-pants for the mother tree scenario but not the sibling plant scenar-io, so we do not report coding analyses for kin altruismexplanations in the main text. Nonetheless, the results from avail-able explanation data roundly supported our overall thesis, asreported here. Mother tree. Among participants who predicted thatthe seedling would grow better closer to its mother (N = 16), wefound that only Ngöbe construed the interaction as one involvingthe mother’s goals for her seedling, as indicated by a reliable cul-tural difference on the 2 � 3 chi-square analysis, v2 (1, N = 16)= 5.88, exact, 2-sided p < 0.05 (using the N � 1 chi-square test).Ngöbe were more likely to refer to social goals than either individ-ual goals or biological mechanisms. In contrast, US participantspredominantly explained the mother-tree interaction in terms ofbiological mechanisms or individual goals. Sibling plants: Ngöbeexplanations for sibling sharing were analyzed, and these showedthe same pattern of social goal interpretations as for the other pre-diction scenarios.

C.5. Experiment 2B: supplementary methods and materials

Question probes for social capacity attributions as translatedinto each study language (English, Spanish) are provided below,with accompanying stimuli indicated in brackets.

Signaling. Do you think that plants can communicate with andsend signals to one another?¿Tú crees que las plantas pueden comunicar o mandar señalesentre ellas mismas?[Insect communication illustration]

Kin recognition. Do you think that plants can recognize whotheir kin are?¿Tú crees que las plantas pueden reconocer quiénes son susdescendentes?[Sibling plant illustration]

Mother help. Do you think that a mother tree can help its babytree via roots?¿Crees que la mama puede ayudar él bebe árbol por medio de susraíces?[Mother tree illustration]

Feel other’s pain. [This item was presented after an initial ques-tion, ‘‘Can this tree feel pain when you cut it?”] Now supposethat there are other trees from the same species growingaround this fallen tree. Can they feel the tree’s pain even thoughthey themselves were not cut?

Ahora supone que hay otros árboles de la misma especie creciendoalrededor de este árbol que ha caído. ¿Ellos pueden sentir el dolordel árbol, aunque ellos mismos no fueron cortados?[Photograph of a chainsaw-fallen tree and stump]

C.6. Experiment 2B: supplementary results

In addition to the cultural differences reported in the main text,there were several main effects and interactions, as would beexpected (evident in Fig. 5 presented in the main text). Maineffects: There was a main effect of capacity, F(3,126) = 8.43,p < 0.001, g2 = 0.17. Plant signaling was most often endorsed(M = 0.81, SD = 0.44) and reliably more so than either feel other’spain (M = 0.41, SD = 0.31) or kin recognition (M = 0.56, SD = 0.50),


ps < 0.05. The capacity for a mother to help offspring (M = 0.66,SD = 0.50) was more often endorsed than feel other’s pain,p < 0.05. Capacity attributions for feel other’s pain and kin recogni-tion did not reliably differ.

Appendix D. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.cognition.2017.01.023.

References

Abramson, C. I., & Chicas-Mosier, A. M. (2016). Learning in plants: Lessons fromMimosa pudica. Frontiers in Psychology, 7.

Alpi, A., Amrhein, N., Bertl, A., Blatt, M. R., Blumwald, E., Cervone, F., ... Galston, A. W.(2007). Plant neurobiology: No brain, no gain? Trends in Plant Science, 12(4),135–136.

Anders, R. (2013). CCTpack: Cultural consensus theory applications to data. R packageversion 1.2.

Anders, R. (2014). Package ‘‘CCTpack” manual Retrieved from <http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.649.4064&rep=rep1&type=pdf>.

Anders, R., Oravecz, Z., & Batchelder, W. H. (2014). Cultural consensus theory forcontinuous responses: A latent appraisal model for information pooling. Journalof Mathematical Psychology, 61, 1–13.

Anderson, E. N. (1996). Ecologies of the heart. New York: Oxford University Press.Anderson, E. N. (2005). Political ecology in a Yucatec Maya community. University of

Arizona Press.Anderson, E. N. (2012). Religion in conservation and management: A Durkheimian

view. Journal for the Study of Religion, Nature and Culture, 6(4), 398–420.Atran, S., & Medin, D. L. (2008). The native mind and the cultural construction of

nature. Cambridge: MIT Press.Atran, S., Medin, D. L., & Ross, N. O. (2005). The cultural mind: Environmental

decision making and cultural modeling within and across populations.Psychological Review, 112(4), 744.

Atran, S., & Norenzayan, A. (2004). Religion’s evolutionary landscape:Counterintuition, commitment, compassion, communion. Behavioral and BrainSciences, 27(06), 713–730.

Bang, M., Medin, D. L., & Atran, S. (2007). Cultural mosaics and mental models ofnature. Proceedings of the National Academy of Sciences, 104(35), 13868–13874.

Barrett, J. L. (2000). Exploring the natural foundations of religion. Trends in CognitiveSciences, 4(1), 29–34.

Bhatt, M. V., Khandelwal, A., & Dudley, S. A. (2011). Kin recognition, not competitiveinteractions, predicts root allocation in young Cakile edentula seedling pairs.New Phytologist, 189(4), 1135–1142.

Biedrzycki, M. L., & Bais, H. P. (2010). Kin recognition in plants: A mysteriousbehaviour unsolved. Journal of Experimental Botany, erq250.

Bird-David, N. (1999). ‘‘Animism” revisited: Personhood, environment, andrelational epistemology 1. Current Anthropology, 40(S1), S67–S91.

Bíró, S., & Leslie, A. M. (2007). Infants’ perception of goal-directed actions:Development through cue-based bootstrapping. Developmental Science, 10(3),379–398.

Boyer, P. (1996). What makes anthropomorphism natural: Intuitive ontology andcultural representations. Journal of the Royal Anthropological Institute, 83–97.

Boyer, P. (2003). Religious thought and behaviour as by-products of brain function.Trends in Cognitive Sciences, 7(3), 119–124.

Brenner, E. D., Stahlberg, R., Mancuso, S., Vivanco, J., Baluška, F., & Van Volkenburgh,E. (2006). Plant neurobiology: An integrated view of plant signaling. Trends inPlant Science, 11(8), 413–419.

Cajete, G. (2000). Native science: Natural laws of interdependence. Santa Fe, NM: ClearLight Books.

Campbell, I. (2007). Chi-squared and Fisher-Irwin tests of two-by-two tables withsmall sample recommendations. Statistics in Medicine, 26(19), 3661–3675.

Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: Bradford Books.Carey, S. (2009). The origin of concepts. Oxford, UK: Oxford University Press.Chamovitz, D. (2012). What a plant knows: A field guide to the senses. Macmillan.Csibra, G., & Gergely, G. (2013). Teleological understanding of actions. In M. R.

Banaji & S. A. Gelman (Eds.), Navigating the social world (pp. 38–43). New York:Oxford University Press.

Csibra, G., Gergely, G., Bıró, S., Koos, O., & Brockbank, M. (1999). Goal attributionwithout agency cues: The perception of ‘pure reason’in infancy. Cognition, 72(3),237–267.

Cvrcková, F., Zársky, V., & Markoš, A. (2016). Plant studies may lead us to rethink theconcept of behavior. Frontiers in Psychology, 7.

Danziger, E. (2010). On trying and lying: Cultural configurations of Grice’s Maxim ofQuality. Intercultural Pragmatics, 7(2), 199–219.

Danziger, E. (2006). The thought that counts: Interactional consequences of variationin cultural theories of meaning. In N. J. Enfield & S. C. Levinson (Eds.), Roots ofhuman sociality: Culture, cognition and interaction (pp. 259–278).Oxford, UK:Berg.

Danziger, E., & Rumsey, A. (2013). Introduction: From opacity to intersubjectivityacross languages and cultures. Language and Communication, 33(3), 247–250.

Dennett, D. C., & Haugeland, J. (1987). Intentionality. In The Oxford companion to themind.

Descola, P. (1996). Constructing natures. Nature and Society: AnthropologicalPerspectives, 82–102.

Dudley, S. A., & File, A. L. (2007). Kin recognition in an annual plant. Biology Letters, 3,435–438.

Duranti, A. (1988). Intentions, language, and social action in a Samoan context.Journal of Pragmatics, 12(1), 13–33.

Duranti, A. (2008). Further reflections on reading other minds. AnthropologicalQuarterly, 81(2), 483–494.

Erickson, J. E., Keil, F. C., & Lockhart, K. L. (2010). Sensing the coherence of biology incontrast to psychology: Young children’s use of causal relations to distinguishtwo foundational domains. Child Development, 81(1), 390–409.

Falik, O., Mordoch, Y., Quansah, L., Fait, A., & Novoplansky, A. (2011).Rumor has it. . .: Relay communication of stress cues in plants. PLoS ONE, 6(11), e23625.

Fausey, C. M., & Boroditsky, L. (2011). Who dunnit? Cross-linguistic differences ineye-witness memory. Psychonomic Bulletin & Review, 18(1), 150–157.

Fausey, C. M., Long, B. L., Inamori, A., & Boroditsky, L. (2010). Constructing agency:The role of language. Frontiers in Psychology, 1.

FORMABIAP (2008). La vida secreta de las plantas medicinales en los puebloskichwa, kukama-kukamiria y tikuna. In P. d. f. d. m. b. d. l. A. . Peruana (Ed.),Una aproximacion al concocimiento de algunas plantas de uso medicinal en lacomunidad educativa de Zungarococha (pp. 228). Iquitos: Serie: Construyendointerculturalidad.

France, S. L., & Batchelder, W. H. (2015). Maximum likelihood item easiness modelsfor test theory without an answer key. Educational and PsychologicalMeasurement, 75(1), 57–77.

Gelman, S. A. (2009). Learning from others: Children’s construction of concepts.Annual Review of Psychology, 60, 115–140.

Gergely, G., & Csibra, G. (2003). Teleological reasoning in infancy: The naıve theoryof rational action. Trends in Cognitive Sciences, 7(7), 287–292.

Goddard, C. (2003). Thinking across languages and cultures: Six dimensions ofvariation. Cognitive Linguistics, 14(2/3), 109–140.

Goddard, C. (2010). Universals and variation in the lexicon of mental state concepts.In B. C. Malt & P. Wolff (Eds.), Words and the mind: How words capture humanexperience (pp. 72–93). New York: Oxford University Press.

Gordon, B. L. (1982). A panama forest and shore: Natural history and amerindianculture in Bocas del Toro. Pacific Grove, California: Boxwood Press.

Gutheil, G., Vera, A., & Keil, F. C. (1998). Do houseflies think? Patterns of inductionand biological beliefs in development. Cognition, 66(1), 33–49.

Guthrie, S., Agassi, J., Andriolo, K. R., Buchdahl, D., Earhart, H. B., Greenberg, M., ...Sharpe, K. J. (1980). A cognitive theory of religion [and Comments and Reply].Current Anthropology, 181–203.

Hall, M. (2011). Suny series on religion and the environment: Plants as persons: Aphilosophical botany. SUNY Press.

Hallowell, A. I. (1960). Ojibwa ontology, behavior, and world view. Readings inIndigenous Religions, 18–49.

Harvey, G. (2005). Animism: Respecting the living world. Kent Town, South Australia:Wakefield Press.

Hatano, G., & Inagaki, K. (1994). Young children’s naive theory of biology. Cognition,50(1), 171–188.

Heidelberger, M. (2004). Nature from within: Gustav Theodor Fechner and hispsychophysical worldview. University of Pittsburgh Press.

Heil, M., & Karban, R. (2010). Explaining evolution of plant communication byairborne signals. Trends in Ecology & Evolution, 25(3), 137–144.

Herrmann, P., Waxman, S. R., & Medin, D. L. (2010). Anthropocentrism is not thefirst step in children’s reasoning about the natural world. Proceedings of theNational Academy of Sciences, 107(22), 9979–9984.

Hirschfeld, L. A. (2013). The myth of mentalizing and the primacy of folk sociology.In M. Banaji & S. A. Gelman (Eds.), Navigating the social world: What infants,children, and other species can teach us (pp. 101–106). Oxford, UK: OxfordUniversity Press.

Hirschfeld, L. A., & Gelman, S. A. (1994). Mapping the mind: Domain specificity incognition and culture. Cambridge University Press.

Hirschfeld, L. A. (2006). Who needs a theory of mind? In R. Viale, D. Andler, & L. A.Hirschfeld (Eds.), Biological and cultural bases of human inference (pp. 131–160).Mahwah, NJ: Lawrence Erlbaum Publishers.

Inagaki, K., & Hatano, G. (2004). Vitalistic causality in young children’s naivebiology. Trends in Cognitive Sciences, 8(8), 356–362.

Ingold, T. (2006). Rethinking the animate, re-animating thought. Ethnos, 71(1),9–20.

Ingold, T. (2010). The man in the machine and the self-builder. InterdisciplinaryScience Reviews, 35(3–4), 353–364.

Ingold, T. (2011). Being alive: Essays on movement, knowledge and description. Taylor& Francis.

Johannessen, C. L. (1970). The dispersal of Musa in Central America: Thedomestication process in action. Annals of the Association of AmericanGeographers, 60(4), 689–698.

Johnson, H. M. (1945). Are psychophysical problems genuine or spurious? TheAmerican Journal of Psychology, 58(2), 189–211.

Johnson, S. C. (2000). The recognition of mentalistic agents in infancy. Trends inCognitive Sciences, 4(1), 22–28.

Johnson, S. C. (2003). Detecting agents. Philosophical Transactions of the Royal Societyof London. Series B: Biological Sciences, 358(1431), 549–559.



http://refhub.elsevier.com/S0010-0277(17)30032-X/h0005





http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.649.4064%26rep=rep1%26type=pdf

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.649.4064%26rep=rep1%26type=pdf




















































































































































Johnson, S. C., Shimizu, Y., & Ok, S.-J. (2007). Actors and actions: The role ofagent behavior in infants’ attribution of goals. Cognitive Development, 22(3),310–322.

Keil, F. C. (1995). The growth of causal understandings of natural kinds: Modes ofconstrual and the emergence of biological thought. In D. Sperber, D. Premack, &A. J. Premack (Eds.), Causal cognition: A multidisciplinary debate (pp. 234–267).Oxford, UK: Clarendon Press.

Kelemen, D. (2003). British and American children’s preferences for teleo-functionalexplanations of the natural world. Cognition, 88(2), 201–221.

Keller, E. F. (1983). A feeling for the organism: The life and work of Barbara McClintock.New York: WH Freeman and Company.

Kimmerer, R. W. (2013). Braiding sweetgrass: Indigenous wisdom, scientific knowledgeand the teachings of plants. Minneapolis, MN: Milkweed Editions.

Klemens, J. A. (2008). Kin recognition in plants? Biology Letters, 4(1), 67–68.Kohn, E. (2007). How dogs dream: Amazonian natures and the politics of

transspecies engagement. American Ethnologist, 34(1), 3–24.Kohn, E. (2013). How forests think: Toward an anthropology beyond the human. Univ

of California Press.Kohn, E. (2015). Anthropology of ontologies. Annual Review of Anthropology, 44,

311–327.Kopesec, M. (1975). Los elementos verbales y sustantivos y la oración en guaymí.

Lenguas de Panamá, 2, 19–82.Le Guen, O., Iliev, R., Lois, X., Atran, S., & Medin, D. L. (2013). A garden experiment

revisited: Inter-generational change in environmental perception andmanagement of the Maya Lowlands, Guatemala. Journal of the RoyalAnthropological Institute, 19(4), 771–794.

Lewis, M. P. (Ed.). (2009). Ethnologue: Languages of the world (16th ed.. Dallas, TX:SIL International.

Lillard, A. (1998). Ethnopsychologies: Cultural variations in theories of mind.Psychological Bulletin, 123, 3–32.

Lima, T. S. (1999). The two and its many: Reflections on perspectivism in a Tupicosmology. Ethnos, 64(1), 107–131.

Lindsay, W. L. (1876). Mind in plants. The British Journal of Psychiatry, 21(96),513–532.

Lowder, M. W., & Gordon, P. C. (2015). Natural forces as agents: Reconceptualizingthe animate–inanimate distinction. Cognition, 136, 85–90.

Luhrmann, T. (2011). Toward an anthropological theory of mind (overview). SuomenAntropologi: Journal of the Finnish Anthropological Society, 36, 5–69.

Mancuso, S., & Viola, A. (2015). Brilliant green: The surprising history and science ofplant intelligence. Island Press.

Marder, M. (2012). Plant intentionality and the phenomenological framework ofplant intelligence. Plant Signaling & Behavior, 7(11), 1365–1372.

Marder, M. (2013). Plant intelligence and attention. Plant Signaling & Behavior, 8(5),e23902. http://dx.doi.org/10.4161/psb.23902.

Medin, D. L., & Bang, M. (2014). Who’s asking? Native science, western science andscience education. Cambridge, MA: MIT Press.

Medin, D. L., ojalehto, b., Waxman, S. R., & Bang, M. (2015). Relations: Language,epistemologies, categories and concepts. In E. Margolis & S. Laurence (Eds.), Theconceptual mind: New directions in the study of concepts. Cambridge, MA: MITPress.

Medin, D. L., Ross, N., Cox, D., & Atran, S. (2007). Why folkbiology matters:Resource conflict despite shared goals and knowledge. Human Ecology, 35(3),315–329.

Meltzoff, A. N., & Gopnik, A. (2013). Learning about the mind from evidence:Children’s development of intuitive theories of perception and personality. In S.Baron-Cohen, H. Tager-Flausber, & M. Lombardo (Eds.), Understanding otherminds (3rd ed., pp. 19–34). Oxford, England: Oxford University Press.

Montezuma, L. P. (1991). Nun Roae Kugwe Kira Niebare Nuen: Nuestros Abuelos NosContaron Historias (Our Grandparents Told Us These Stories). Comarca Ngobe,Panama: Accion Cultural Ngobe.

Narby, J. (2006). Intelligence in nature. In An inquiry into knowledge. Penguin. com.ojalehto, b., Medin, D., & Garcia, S. (2015). Agency concepts across cultures: How

intuitive is folkpsychology? In Paper presented at the 37th annual meeting of thecognitive science society, Pasadena, CA.

ojalehto, b., Medin, D., & Garcia, S. (2017). Grounding principles for inferringagency: Two cultural perspectives (submitted for publication).

ojalehto, b., & Medin, D. L. (2015). Perspectives on culture and concepts. AnnualReview of Psychology, 66, 249–275. http://dx.doi.org/10.1146/annurev-psych-010814-015120.

ojalehto, b., Medin, D., Horton, W., Garcia, S., & Kays, E. (2015b). Seeing cooperationor competition: Ecological interactions in cultural perspectives. Topics inCognitive Science (Anthropological perspectives on cognition), 1–22. http://dx.doi.org/10.1111/tops.12156.

ojalehto, b., Waxman, S. R., & Medin, D. L. (2013). Teleological reasoning aboutnature: Intentional design or relational perspectives? Trends in CognitiveSciences, 17(4), 166–171.

Opfer, J. E. (2002). Identifying living and sentient kinds from dynamic information:The case of goal-directed versus aimless autonomous movement in conceptualchange. Cognition, 86(2), 97–122.

Opfer, J. E., & Gelman, S. A. (2001). Children’s and adults’ models for predictingteleological action: The development of a biology-based model. ChildDevelopment, 72(5), 1367–1381.

Opfer, J. E., & Siegler, R. S. (2004). Revisiting preschoolers’ ‘‘living things” concept: Amicrogenetic analysis of conceptual change in basic biology. Cognitivepsychology, 49(4), 301–332.

Pierotti, R. (2011). Indigenous knowledge, ecology, and evolutionary biology. NewYork: Routledge: Taylor & Francis Group.

Pitman, N. C., Cecilio, M. P., Pudicho, M. P., Graham, J. G., Núñez, V. M. P.,Valenzuela, M., & Terborgh, J. W. (2011). Indigenous perceptions of treespecies abundance across an upper Amazonian landscape. Journal ofEthnobiology, 31(2), 233–243.

Ramos, A. R. (2012). The politics of perspectivism. Annual Review of Anthropology, 41,481–494.

Rappaport, R. A. (1979). Ecology, meaning, and religion. Richmond, CA: North AtlanticBooks.

Reichel-Dolmatoff, G. (1976). Cosmology as ecological analysis: A view from therain forest. Man, 307–318.

Reid, A. (1994). Diccionario Guaymi-Espanol-Ingles. Diccionario Guaymi-Espanol-Ingles. Panama: Asamblea Espiritual Nacional de los Baha’is de Panama.

Robbins, J., & Rumsey, A. (2008). Introduction: Cultural and linguistic anthropologyand the opacity of other minds. Anthropological Quarterly, 81(2), 407–420.

Romney, A. K.,Weller, S. C., & Batchelder,W.H. (1986). Culture as consensus: A theoryof culture and informant accuracy. American Anthropologist, 88(2), 313–338.

Ross, N., Medin, D., Coley, J. D., & Atran, S. (2003). Cultural and experientialdifferences in the development of folkbiological induction. CognitiveDevelopment, 18(1), 25–47.

Santos-Granero, F. (2009). The occult life of things: Native Amazonian theories ofmateriality and personhood. University of Arizona Press.

Santos-Granero, F. (2012). ‘‘Here No Poop”: The dangers and prospects ofintersubjective relations in Amerindian political economies of life.Anthropology and Humanism, 37(2), 134–145.

Sarsaneda del Cid, J. (2009). Ni Ngóbe Tó Blitde Ño: Cómo Hablan Los Ngóbe. Panama:Acción Cultural Ngóbe.

Simard, S. W. (2009). The foundational role of mycorrhizal networks in self-organization of interior Douglas-fir forests. Forest Ecology and Management, 258(S), S95–S107.

Sperber, D., & Hirschfeld, L. A. (2004). The cognitive foundations of cultural stabilityand diversity. Trends in Cognitive Sciences, 8(1), 40–46.

Trewavas, A. (2002). Plant intelligence: Mindless mastery. Nature, 415(6874), 841.Trewavas, A. (2003). Aspects of plant intelligence. Annals of Botany, 92(1), 1–20.Trewavas, A. (2016). Intelligence, cognition, and language of green plants. Frontiers

in Psychology, 7.Uexküll, J. v. (2010). A foray into the worlds of animals and humans: With a theory

of meaning. In J. D. O’Neill (Ed.). Trans. (Vol. 12). U of Minnesota Press.Unsworth, S. J., Levin, W., Bang, M., Washinawatok, K., Waxman, S. R., & Medin, D. L.

(2012). Cultural differences in childrens ecological reasoning and psychologicalcloseness to nature: Evidence from Menominee and European Americanchildren. Journal of Cognition and Culture, 12(1–2), 17–29.

Viveiros de Castro, E. (1998). Cosmological deixis and Amerindian perspectivism.Journal of the Royal Anthropological Institute, 4(3), 469–488.

von Wiesner, J. (1878). Die Heliotropischen Erscheinungen im Pflanzenreiche: EinePhysiologische Monographie. [Secondary source, cited from Whippo &Hangarter, 2006]. Kaiserlich-ko niglichen Hof- und Staatsdruckerei.

Waldstein, A., & Adams, C. (2006). The interface between medical anthropology andmedical ethnobiology. Journal of the Royal Anthropological Institute, 12(s1),S95–S118.

Waxman, S. R., & Medin, D. L. (2007). Experience and cultural models matter:Placing firm limits on childhood anthropocentrism. Human Development, 50(1),23–30.

Weaver, B. (2013). Equivalence of Linear-by-Linear Chi-Square and the ’N-1’ chi square,2015. From <sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics>.

Wellman, H. M. (2010). Developing a theory of mind. The Blackwell handbook ofcognitive development, 258–284.

Wellman, H. M., & Gelman, S. A. (1992). Cognitive development: Foundationaltheories of core domains. Annual Review of Psychology, 43(1), 337–375.

Wellman, H. M., & Johnson, C. N. (2008). Developing dualism: From intuitiveunderstanding to transcendental ideas. In A. Antonietti, A. Corradina, & E. Lowe(Eds.), Psycho-physical dualism today: An interdisciplinary approach (pp. 3–36).Lanham, MD: Lexington Books.

Wertz, A. E., & Wynn, K. (2014a). Selective social learning of plant edibility in 6-and18-month-old infants. Psychological Science. 0956797613516145.

Wertz, A. E., & Wynn, K. (2014b). Thyme to touch: Infants possess strategies thatprotect them from dangers posed by plants. Cognition, 130(1), 44–49.

Whippo, C. W., & Hangarter, R. P. (2006). Phototropism: Bending towardsenlightenment. The Plant Cell Online, 18(5), 1110–1119.

Wilkinson, D. M. (1998). The evolutionary ecology of mycorrhizal networks. Oikos,82(2), 407–410.

Woodward, A. L. (2013). Infant foundations of intentional understanding. In M. R.Banaji & S. A. Gelman (Eds.), Navigating the social world: What infants, children,and other species tell us (pp. 75–80). Oxford University Press.

Young, P. D. (1975). Guaymi nativism: its rise and demise. In Paper presented at theXLI international congress of Americanists.

Young, P. D. (1986). Ngawbere numeral classifiers. In Paper presented at the secondannual meeting of the Pacific linguistics conference, Eugene, Oregon.











































http://dx.doi.org/10.4161/psb.23902

















http://dx.doi.org/10.1146/annurev-psych-010814-015120

http://dx.doi.org/10.1146/annurev-psych-010814-015120

http://dx.doi.org/10.1111/tops.12156

http://dx.doi.org/10.1111/tops.12156


































































http://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics




















conceptualizing agency: folkpsychological and ... agency: folkpsychological and folkcommunicative...

Documents